U.S. patent application number 13/569203 was filed with the patent office on 2012-11-22 for appapatus and method for generating a collection profile and for communicating based on the collection profile.
This patent application is currently assigned to BACH TECHNOLOGY AS. Invention is credited to Dagfinn BACH, Sebastian SCHMIDT.
Application Number | 20120296908 13/569203 |
Document ID | / |
Family ID | 41165520 |
Filed Date | 2012-11-22 |
United States Patent
Application |
20120296908 |
Kind Code |
A1 |
BACH; Dagfinn ; et
al. |
November 22, 2012 |
APPAPATUS AND METHOD FOR GENERATING A COLLECTION PROFILE AND FOR
COMMUNICATING BASED ON THE COLLECTION PROFILE
Abstract
An apparatus for generating a collection profile of a collection
of different media data items has a feature extractor for
extracting at least two different features describing a content of
a media data item for a plurality of media data items of the
collection, and a profile creator for creating the collection
profile by combining the extracted features or weighted extracted
features so that the collection profile represents a quantitative
fingerprint of a content of the media data collection. This
collection profile or music DNA can be used for transmitting
information, which is based on this collection profile, to the
entity itself or to a remote entity.
Inventors: |
BACH; Dagfinn; (Bergen,
NO) ; SCHMIDT; Sebastian; (Ilmenau, DE) |
Assignee: |
BACH TECHNOLOGY AS
Bergen
NO
|
Family ID: |
41165520 |
Appl. No.: |
13/569203 |
Filed: |
August 8, 2012 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13035083 |
Feb 25, 2011 |
|
|
|
13569203 |
|
|
|
|
PCT/EP2009/006042 |
Aug 20, 2009 |
|
|
|
13035083 |
|
|
|
|
Current U.S.
Class: |
707/737 ;
707/E17.089 |
Current CPC
Class: |
G06F 16/683 20190101;
G06F 16/637 20190101; G06F 16/639 20190101; G06F 16/68
20190101 |
Class at
Publication: |
707/737 ;
707/E17.089 |
International
Class: |
G06F 17/30 20060101
G06F017/30 |
Foreign Application Data
Date |
Code |
Application Number |
Aug 28, 2008 |
EP |
08015229.1 |
Dec 18, 2008 |
EP |
08022058.5 |
Claims
1. An apparatus for generating a collection profile of a collection
of different media data items, the media items being audio files,
comprising: a feature extractor configured to extract at least two
different features describing a content of a media data item from a
plurality of media data items of the collection; and a profile
creator configured to create the collection profile by combining
the extracted features or weighted extracted features for the
plurality of media data items so that the collection profile
represents a quantitative fingerprint of a content of the
collection, wherein the apparatus further comprises an input
configured to receive information on a music taste of a user of the
collection of different audio files, and wherein the profile
creator is configured to create a raw collection profile without
information on a user behavior logged by the profile creator or
information on a music taste, and to weight the raw collection
profile using weights derived from the information on the music
taste or the user behavior to acquire the collection profile,
wherein at least one of the feature extractor, the profile creator,
and the input comprises a hardware implementation.
2. The apparatus in accordance with claim 1, in which the profile
creator is configured to generate the quantitative collection
profile using an addition, a subtraction, a multiplication or a
division of the extracted features or the weighted extracted
features.
3. The apparatus in accordance with claim 1, in which the
collection is stored on a mobile device comprising a storage for
media data items; and in which the profile creator is configured to
add information on a manufacturer of the mobile device or
information on a specific model of the mobile device or information
on a communication service provider to the collection profile.
4. The apparatus in accordance with claim 1, in which each media
data item of the plurality of media data items comprises a data
portion and an associated metadata portion, the metadata portion
comprising the at least two different features, and in which the
feature extractor is configured to parse the metadata portion and
to read the features from the metadata portion.
5. The apparatus in accordance with claim 1, in which the features
describing the content comprise information on energy-strong,
energy-mid, energy-low, color-bright, color-dark, tempo/beat,
percussiveness, music color, vocal detection, aggressiveness,
segmentation, key, mood, solo instrument or fingerprint.
6. The apparatus in accordance with claim 1, in which the profile
creator is configured to analyze the extracted features for similar
features or to exclude features comprising extreme values.
7. The apparatus in accordance with claim 1, in which the feature
extractor is configured to detect a newly stored media data item in
the collection, and in which the profile creator is configured to
retrieve an earlier collected profile and to manipulate the earlier
collected profile based on the features of the newly stored media
data item to create an updated collection profile.
8. An apparatus for generating a collection profile of a collection
of different media data items, the media items being audio files,
comprising: a feature extractor configured to extract at least two
different features describing a content of a media data item from a
plurality of media data items of the collection; and a profile
creator configured to create the collection profile by combining
the extracted features or weighted extracted features for the
plurality of media data items so that the collection profile
represents a quantitative fingerprint of a content of the
collection, wherein the profile creator is configured to weight the
extracted features using weights derived from the information on
the music taste or the user behavior logged by the profile creator,
and to combine the weighted extracted features to acquire the
collection profile, wherein at least one of the feature extractor,
the profile creator, and the input comprises a hardware
implementation.
9. An apparatus for communicating information from a first entity
to a second entity, comprising: a collection profile information
generator configured to generate information on a first collection
profile of a first collection at the first entity or for generating
a second collection profile of a second collection at the second
entity, the first or second collection profile representing a
quantitative fingerprint of a content of the corresponding
collection of metadata files at the first entity or at the second
entity; and an information handler configured to use the first
collection profile or the second collection profile within a
matching operation and for transmitting information based on the
matching operation or a collection profile or for receiving a
message based on a matching operation or the collection profile,
wherein at least one of the collection profile information
generator, and the information handler comprises a hardware
implementation.
10. The apparatus in accordance with claim 9, in which the
apparatus is a mobile device representing the first entity, in
which the apparatus is configured to generate the first collection
profile, and in which the information handler is configured to send
a message comprising information on the first collection profile to
the second entity.
11. The apparatus in accordance with claim 9, where the apparatus
is a device representing the first entity, and where the apparatus
is configured to determine information on the collection profile,
in which the message handler is configured to retrieve information
on a user group comprising users with a similar collection profile
as the first collection profile and to send a message to the second
entity requesting a joining to a user group, or to retrieve
information on a user comprising a similar profile and to send a
message to the second user, the message comprising information on
the first entity or a media data item from the collection, or to
retrieve information on an advertisement tailored to a product or
service related to a collection profile and to send a message to
the advertiser requesting advertisements tailored to the collection
profile or to deliver an advertisement for a product to a different
entity, or to remotely analyze stored media data items for a
matching profile and to selectively request a download of a media
data item comprising the matching profile.
12. The apparatus in accordance with claim 9, in which the
collection profile information generator is configured to receive,
via cable or via a wireless communication device, information on
profiles of a plurality of entities located in a limited space
around the second entity or being in a network user group, in which
the information handler is configured to calculate a combined
profile for the plurality of first entities, in which the
information handler is configured to select a media data item
matching with the combined profile, and in which the information
handler is configured to cause a rendering of the selected media
data item in the limited space or to cause a data streaming of the
selected media data item to the network user group.
13. The apparatus in accordance with claim 9, in which the
information handler is configured to calculate a result list from
the matching operation, to map confidence measures for a selected
group of media items from the result list, and to use a final
result list comprising the mapped confidence measures for
transmitting data to the user.
14. The apparatus in accordance with claim 9, wherein the
collection profile information generator is configured to retrieve
the collection profile of the first entity, wherein the information
handler is configured to modify the collection profile based on a
current taste of a user of the first entity or based on an external
situation, wherein the information handler is configured to perform
the matching operation with the modified profile in the media
collection of the first entity, and wherein the information handler
is configured to play a result of the matching operation at the
first entity or to generate a play list at the first entity based
on the result of the matching operation.
15. A method of generating a collection profile of a collection of
different media date items, the media items being audio files,
comprising: extracting at least two different features describing a
content of a media date item from a plurality of media data items
of the collection; and creating, by a profile creator, the
collection profile by combining the extracted features or weighted
extracted features for the plurality of media date items so that
the collection profile represents a quantitative fingerprint of a
content of the collection, wherein the method further comprises
receiving, by an input, information on a music taste of a user of
the collection of different audio files, and wherein creating
comprises: creating a raw collection profile without information on
a user behavior logged by the profile creator or information on a
music taste, and weighting the raw collection profile using weights
derived from the information on the music taste or the user
behavior to acquire the collection profile.
16. A method of generating a collection profile of a collection of
different media date items, the media items being audio files,
comprising: extracting at least two different features describing a
content of a media date item from a plurality of media data items
of the collection; and creating, by a profile creator, the
collection profile by combining the extracted features or weighted
extracted features for the plurality of media date items so that
the collection profile represents a quantitative fingerprint of a
content of the collection, wherein creating comprises weighting the
extracted features using weights derived from the information on
the music taste or the user behavior logged by the profile creator,
and combining the weighted extracted features to acquire the
collection profile.
17. A method of communicating information from a first entity to a
second entity, comprising: generating information on a first
collection profile of a first collection at the first entity or
generating a second collection profile of a second collection at
the second entity, the collection profile representing a
quantitative fingerprint of a content of the corresponding
collection of metadata files at the first entity or at the second
entity; using the first collection profile or the second collection
profile within a matching operation; and transmitting information
based on the matching operation or a collection profile or
receiving a message based on a matching operation or the collection
profile.
18. A non-transitory storage medium having stored thereon a
computer program comprising a program code for performing a method
in accordance with claim 15, when running on a computer.
19. A non-transitory storage medium having stored thereon a
computer program comprising a program code for performing a method
in accordance with claim 16, when running on a computer.
20. A non-transitory storage medium having stored thereon a
computer program comprising a program code for performing a method
in accordance with claim 17, when running on a computer
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is a continuation of copending U.S.
Non-Provisional application Ser. No. 13/035,083, filed Feb. 25,
2011, which is a Continuation of International Application No.
PCT/EP2009/006042, filed Aug. 20, 2009, which is incorporated
herein by reference in its entirety, and additionally claims
priority from U.S. Provisional Application No. 61/092,528, filed
Aug. 28, 2008 and from European Applications Nos. 08015229.1, filed
Aug. 28, 2008, and 08022058.5, filed Dec. 18, 2008, which are all
incorporated herein by reference in their entirety.
BACKGROUND OF THE INVENTION
[0002] The present invention is related to media data processing
and, particularly, to media data characterization and usage
thereof.
[0003] Mobile media data players, such as MP3 players or MP4
players, are becoming more and more popular. Furthermore, due to
the high data compression rates obtained by modern audio coding
tools, the storage requirements for the individual audio or video
files are decreasing. Concurrently, the prices for memories are
decreasing as well. These memories can be chip memories or hard
disks, which are used in mobile video or audio players. Since the
prices for these memories drop more and more, users can store more
and more different media data items on their players.
[0004] Furthermore, audio and/or video playing functionalities are
not restricted to dedicated audio or video players, but are
available in even more mobile devices, such as mobile phones, PDAs,
navigation devices, etc. Naturally, notebook computers also have
full audio/video-playing functionalities and, of course, almost
unlimited storage resources due to huge hard disk resources.
[0005] In view of that, users store more and more media data items,
such as audio files or video files, on their electronic data
carriers, and the task for managing these increasing databases is
become more and more difficult. Many mobile devices support the ID3
tag, which indicates, for an audio file, the title of the audio
file, the author or a band performing the audio file, an album to
which the audio file belongs to and, for example, the playing time
of the audio file. These information items can be read by a mobile
player, can be displayed by the mobile player and can be used for
assisting the user in editing playlists, etc.
[0006] Furthermore, the MPEG-7 standard has defined the inclusion
of additional metadata to audio files. These audio file metadata
include additional features related to the audio content of a media
data item, such as tempo, beats per minute, etc. These features can
be used for characterizing a media data item in a much more
content-related way and these features form a basis for finally
reaching a search capability among audio files comparable to search
functionalities in text files.
[0007] The specific selection of a media data item on an electronic
storage device represents a very user-specific individual
collection, since each user will load different media data items on
her or his specific personal electronic storage device. Therefore,
a collection of media data items will be a very personal issue
characterizing the user of this electronic media data storage
device.
SUMMARY
[0008] According to an embodiment, an apparatus for generating a
collection profile of a collection of different media data items,
the media items being audio files, may have: a feature extractor
for extracting at least two different features describing a content
of a media data item from a plurality of media data items of the
collection; and a profile creator for creating the collection
profile by combining the extracted features or weighted extracted
features for the plurality of media data items so that the
collection profile represents a quantitative fingerprint of a
content of the collection, wherein the apparatus further has an
input for receiving information on a music taste of a user of the
collection of different audio files, and wherein the profile
creator is operative to create a raw collection profile without
information on a user behavior logged by the profile creator or
information on a music taste, and to weight the raw collection
profile using weights derived from the information on the music
taste or the user behavior to obtain the collection profile.
[0009] According to another embodiment, an apparatus for generating
a collection profile of a collection of different media data items,
the media items being audio files, may have: a feature extractor
for extracting at least two different features describing a content
of a media data item from a plurality of media data items of the
collection; and a profile creator for creating the collection
profile by combining the extracted features or weighted extracted
features for the plurality of media data items so that the
collection profile represents a quantitative fingerprint of a
content of the collection, wherein the profile creator is operative
to weight the extracted features using weights derived from the
information on the music taste or the user behavior logged by the
profile creator, and to combine the weighted extracted features to
obtain the collection profile.
[0010] According to another embodiment, an apparatus for
communicating information from a first entity to a second entity
may have: a collection profile information generator for generating
information on a first collection profile of a first collection at
the first entity or for generating a second collection profile of a
second collection at the second entity, the first or second
collection profile representing a quantitative fingerprint of a
content of the corresponding collection of metadata files at the
first entity or at the second entity, the first or second
collection profile being generated by an apparatus for generating a
collection profile as mentioned above; and an information handler
for using the first collection profile or the second collection
profile within a matching operation and for transmitting
information based on the matching operation or a collection profile
or for receiving a message based on a matching operation or the
collection profile.
[0011] According to another embodiment, a method of generating a
collection profile of a collection of different media date items,
the media items being audio files, may have the steps of:
extracting at least two different features describing a content of
a media date item from a plurality of media data items of the
collection; and creating, by a profile creator, the collection
profile by combining the extracted features or weighted extracted
features for the plurality of media date items so that the
collection profile represents a quantitative fingerprint of a
content of the collection, wherein the method further has the step
of receiving, by an input, information on a music taste of a user
of the collection of different audio files, and wherein the step of
creating has: creating a raw collection profile without information
on a user behavior logged by the profile creator or information on
a music taste, and weighting the raw collection profile using
weights derived from the information on the music taste or the user
behavior to obtain the collection profile.
[0012] According to still another embodiment, a method of
generating a collection profile of a collection of different media
date items, the media items being audio files, may have the steps
of: extracting at least two different features describing a content
of a media date item from a plurality of media data items of the
collection; and creating, by a profile creator, the collection
profile by combining the extracted features or weighted extracted
features for the plurality of media date items so that the
collection profile represents a quantitative fingerprint of a
content of the collection, wherein the step of creating has
weighting the extracted features using weights derived from the
information on the music taste or the user behavior logged by the
profile creator, and combining the weighted extracted features to
obtain the collection profile.
[0013] According to another embodiment, a method of communicating
information from a first entity to a second entity may have the
steps of: generating information on a first collection profile of a
first collection at the first entity or generating a second
collection profile of a second collection at the second entity, the
collection profile representing a quantitative fingerprint of a
content of the corresponding collection of metadata files at the
first entity or at the second entity, the first or second
collection profile being generated by a method for generating a
collection profile as mentioned above; using the first collection
profile or the second collection profile within a matching
operation; and transmitting information based on the matching
operation or a collection profile or receiving a message based on a
matching operation or the collection profile.
[0014] Another embodiment may have a computer program having a
program code for performing a method of generating a collection
profile as mentioned above or a method of communicating information
as mentioned above, when running on a computer.
[0015] The present invention is based on the finding that the very
user-specific and, specifically, person-specific collection of
media data items on an electronic storage device or, generally, a
collection of different media data items owned by a certain user
can be advantageously utilized for the purpose of characterizing
the owner of the collection of different media data, or, generally,
the user of an electronic storage device having stored thereon a
collection of different media data. In a first phase, different
features describing a content of a media data item are extracted
for a plurality of media data items. In a second phase, a profile
creator creates the collection profile by combining the extracted
features or by combining weighted (e.g. multiplied by a factor
different from 1 and different from zero) extracted features of the
plurality of media data items so that a quantitative collection
profile is obtained, which represents a quantitative fingerprint of
a content of the collection.
[0016] This quantitative fingerprint, when generated based on audio
data, can also be called a "music DNA", since this music DNA
individually characterizes the collection of different media data
items. Since a collection of different media data items will be
different for each person having generated a collection of
different media data items, the quantitative fingerprint for each
media data collection will be different from quantitative
fingerprints from other media data items or other media item
collections.
[0017] Generally, the inventive collection profile, which is a
quantitative fingerprint, fulfills two different and contradicting
requirements for a fingerprint. On the one hand, the data amount
needed by the fingerprint should be as small as possible so that
the storage amount needed for storing a fingerprint is not too high
and, even more important, the processing requirements for a usage
of this fingerprint within a database, where this fingerprint is to
be matched with other corresponding fingerprints to find similar
profiles or fingerprints, are the smaller the smaller the
fingerprint is.
[0018] On the other hand, the fingerprint has to be as
characteristic as possible for the item indicated by the
fingerprint. To make sure that the fingerprint is characteristic
for a user collection, the fingerprint is not derived, for example,
by a pure hashing or any other highly lossy compression, but the
fingerprint is derived from features representing a content of the
media data so that the fingerprint does not represent just the
media data files without any further information, but the
fingerprint actually represents the content of the media data items
rather than, for example, a waveform or a bit structure of media
data items.
[0019] The inventive collection profile, therefore, provides the
ability to be processed by electronic data equipment due to the
fact that it is a quantitative fingerprint which, due to its
quantitativeness, can be processed within database matching
operations etc. On the other hand, the fingerprint is derived from
useful content features so that the fingerprint represents the
content of a media data collection which, in an embodiment, will be
a kind of an average over each separate feature so that the
fingerprint consists of a collection or a vector of several
averaged different features.
[0020] Importantly, the fingerprint does not indicate a certain
waveform or a certain bit structure, but indicates the media taste,
such as the audio taste or the video taste of the user of the
collection of different media data items.
[0021] In accordance with the present invention, this collection
profile which does not represent a metadata for a single file, but
which represents a metadata for a collection of many different
media data items, can be used for many different purposes which all
have in common that, based on a collection profile and a certain
usage of a collection profile within a database application, a
certain communication operation of an electronic data processing
equipment is conducted so that the user performs a certain action
or receives a certain service optimally suited for her or his media
taste represented by the collection profile, which can, for
example, also be called a music DNA, when the media data items are
audio files.
BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Subsequently, embodiments of the present invention are
explained in detail by referencing the enclosed figures, in
which:
[0023] FIG. 1 illustrates a schematic overview of an apparatus for
generating a collection profile in accordance with an embodiment of
the present invention;
[0024] FIG. 2 illustrates a block diagram of an apparatus for
communicating information from a first entity to a second entity in
a schematic view;
[0025] FIG. 3 illustrates a process for generating an output
collection profile having information on the media collection and,
additionally, information on the storage device storing the media
data collection;
[0026] FIG. 4a illustrates an exemplary processing of different
feature vectors in order to generate a music DNA, and the further
processing of the music DNA for the purpose of matching;
[0027] FIG. 4b illustrates a result list of the matching operation
as obtained from FIG. 4a;
[0028] FIG. 4c illustrates a diagram of the distance versus the
media item number of a sorted result list for illustrating the
mapping of the confidence measure;
[0029] FIG. 4d illustrates a flow chart for illustrating an
embodiment of an inventive matching operation, in which confidence
measures are used;
[0030] FIG. 5a illustrates a certain usage scenario of the
collection profile for the purpose of playing selected audio pieces
for a group or a room;
[0031] FIG. 5b illustrates an alternative implementation scenario,
in which a user can join a user group based on his collection
profile;
[0032] FIG. 5c illustrates an alternative implementation, in which
the music DNA of a user is used for providing him with matching
media items or with advertisements for products or services related
to her or his media data taste;
[0033] FIG. 5d illustrates an alternative implementation scenario,
in which the music DNA of a user is modified in accordance with the
current user taste or an external situation, so that the user can
find a matching media item from his own profile using the modified
collection profile;
[0034] FIG. 6 illustrates a schematic representation of a user
DNA;
[0035] FIG. 7 illustrates a feature overview of qualitative and
quantitative features, which can be combined into a user collection
profile;
[0036] FIG. 8 illustrates a schematic representation of an audio
file having associated metadata;
[0037] FIG. 9 illustrates an explanation of the application
scenario, in which products related to certain collection profiles
are selected and advertised;
[0038] FIG. 10 illustrates a schematic diagram of the apparatus for
generating a collection of different media data items; and
[0039] FIG. 11 illustrates the data format of a media metadata
portion for a media data item.
DETAILED DESCRIPTION OF THE INVENTION
[0040] FIG. 1 illustrates in a schematic way an embodiment of an
apparatus for generating a collection profile 50 of a collection 45
of different media data items 45a, 45b, 45c, 45d. Advantageously,
these media data items are stored within a storage 40 of a mobile
device, wherein this mobile device may comprise a mobile phone, an
audio data player, a video data player, a personal digital
assistant (PDA), a notebook, a navigation device or a similar
device having the capability of storing media data files, such as a
memory stick, which may or may not have any audio data or video
data playing capabilities. Furthermore, the collection may comprise
all media data items on a storage device or may comprise only a
certain portion of media data items on a storage device or may even
comprise the storage elements of several different devices owned by
a user, such as a personal computer, a notebook, a mobile player,
all owned by one and the same user. When, however, electronic
storage devices storing media data items of a user are all
synchronized to each other, then it will be sufficient to have only
a single device for deriving the collection profile in order to
have a very sophisticated user DNA.
[0041] The apparatus comprises a feature extractor 10 for
extracting at least two different features 11 describing a content
of a media data item, for a plurality of media data items of the
collection. Therefore, the feature extractor 10 will process one
media data item after the other in order to derive, for each media
data item, the different features F1, F2, Fi. In one embodiment,
the feature extractor 10 is operative to actually analyze the media
data item to derive the at least two different features describing
the content of the media data item. In other embodiments, the media
data items have already associated metadata as indicated at 45 in
FIG. 1 so that the feature extractor only has to parse and evaluate
the metadata portion of a media data item in order to extract the
at least two different features F1, F2 of a single media data item.
All the features extracted by the feature extractor 10 are supplied
to a profile creator 20 for creating the collection profile. The
profile creator combines the extracted features or weighted
extracted features for the plurality of media data items so that
the collection profile represents a quantitative fingerprint of a
content of the media data collection. When the media data files are
audio files, then the collection profile is the user music DNA,
wherein the term "DNA", which normally means deoxyribonucleic acid
in biology, stands for a very individual user-specific
characterization of the user's music taste.
[0042] The collection profile 50 of the user can be stored for
later use or can be transmitted to a different entity via an
information handler 60, or can also be used for performing matching
operations in a database, where this database may comprise
collection profiles of other users or fingerprints for audio data
items, where a matching audio data item which best matches the
collection profile can be located, etc. Generally, the information
handler 60 will perform an action 65 which, in a way, is based on
the collection profile in order to provide the user of the
collection profile or a different entity, with which the user of
the collection profile may communicate, with a certain
message/service, etc.
[0043] In an embodiment, the profile creator 20 is operative to
generate the quantitative collection profile using an addition, a
subtraction, a multiplication or a division of the extracted
features or of the weighted extracted features. Specifically,
features belonging to the same content are combined among
themselves, so that the collection profile which is obtained in the
end by the profile creator has a number of collection profile
components which is identical to the number of features extracted
by the feature extractor.
[0044] FIG. 2 illustrates an apparatus for communicating
information from a first entity to a second entity in accordance
with an embodiment of the present invention. The apparatus for
communicating comprises a collection profile information generator
70 for generating information on a first collection profile of a
first collection at the first entity or for generating information
on a second collection profile of a second collection at the second
entity. Specifically, the collection profile represents a
quantitative fingerprint of a content of the collection. As
discussed in connection with claim 1, the collection profile
information generator 70 can comprise the same components as the
device in FIG. 1. Alternatively, the collection profile information
generator 70 can retrieve an earlier generated collection profile
from a collection profile storage, where the stored collection
profile can be generated by a device in accordance with FIG. 1,
which is separate from the collection profile information generator
70 of FIG. 2.
[0045] The apparatus illustrated in FIG. 2 furthermore comprises an
information handler, which corresponds to the information handler
60 of FIG. 1, since the device 60 in FIG. 2 receives similar input
as the corresponding element 60 of FIG. 1. The information handler
is operative for transmitting information based on the first
collection profile or the second collection profile to the first
entity or the second entity. Alternatively, the information handler
60 is operative to receive a message based on the first collection
profile or the second collection profile from the first entity or
the second entity depending on whether the apparatus for
communicating is residing in the first entity or the second entity.
Specifically, when the apparatus for communicating illustrated in
FIG. 2 is residing in the first entity, a transmission takes place
to the second entity and a reception takes place from the second
entity. When, however, the apparatus illustrated in FIG. 2 is
residing within the second entity, then a transmission to the first
entity or a reception from the first entity takes place.
[0046] Subsequently, FIG. 3 will be described in more detail in
order to show an advantageous sequence of operations to be
performed in this or a similar order for outputting a collection
profile such as a music DNA.
[0047] In a step 81, features for audio files are extracted. The
operation performed in step 81 corresponds to the procedure
performed by item 10 of FIG. 1. In a step 82, a predefined number
of extreme values are deleted. Therefore, in order to make sure
that any reading or extraction errors do no influence the final
collection profile too much, a number of largest or extreme values
from the extracted features of all audio files are deleted.
[0048] Subsequent to the step of deleting indicated at 82 in FIG.
3, a raw collection profile is calculated as indicated at step 83.
To this end, the corresponding features of the different audio
files are combined to obtain a quantitative raw collection profile
subsequent to step 83. Then, in a step 84, the raw collection
profile or individual features are weighted in accordance with a
user taste input 85. In this embodiment, a user can input some
taste specifics, and this is a way that the user can influence his
collection profile/music DNA. This weighting can take place with
the raw collection profile in a specific way so that a certain
taste corresponds to certain weighting factors for different
features. Alternatively, the individual features can be weighted
before being combined by step 83. Thus, a user can increase the
influence of a certain media data item on the user profile by
applying a weighting factor larger than 1 to an advantageous media
data item and, probably, by applying weighting factors smaller than
1 to all other media data items.
[0049] In a further embodiment, the output of step 84 is refined in
a step 86 based on statistics of a usage behavior which is input
into block 86 via usage behavior input 87. Step 86 makes sure that
the collection profile is "living" in that it changes with the
changing user habits. Again, the profile refinement based on the
statistics of the usage behavior can take place using the raw
collection profile at the output of step 83 or the weighted raw
collection profile at the output of step 84. Alternatively, as
indicated in FIG. 3, the statistics of the usage behavior can be
accounted for by weighting extracted features so that a feature
vector of a certain media data item which is often used is
emphasized with respect to a feature vector of a media data item
which is used less times. Additionally, the time from the last
usage to the actual time can be accounted for so that the fact that
a media data item has been used more recently leads to an
emphasizing of this media data item with respect to media data
items which have been used not so recently.
[0050] The output of step 86 is a specific up-to-date and
user-taste-adapted collection profile, which may be improved by
adding information on a mobile device and/or a communication
service provider as indicated in step 87. Specifically, information
on the manufacturer of the device to which the electronic storage
device is operatively connected, such as the manufacturer of the
mobile phone or the manufacturer of the notebook is added to the
user collection and, additionally, information on the communication
provider may be added as well in order to provide an output
collection profile 89 which is useful for many applications which
will be described with respect to FIGS. 5a to 5c.
[0051] In a different embodiment, the output of step 83 can be used
as it is without any of the additional steps 84, 86, 88.
Furthermore, only one or two of the steps 84, 86, 88 can be applied
to the output of step 83 in order to provide an output collection
profile.
[0052] In a further embodiment, the generation of the output
collection profile/user DNA is conducted in an automatic way. To
this end, an extraction of suitable DNA tags (metadata) from all
songs in the collection is performed. Then, the tags are analyzed
for similar features and outliers are excluded. Then, the
high-level tags are weighted based on the user taste, which
advantageously corresponds to the alternative in FIG. 3, in which
the individual feature vectors are weighted before being combined.
Additionally, the profile is refined by using statistics of the
usage behavior. Apart from a time usage, this feature also allows
to prefer certain moods to other less advantageous moods.
[0053] Subsequently, an exemplary determination of the collection
profile is discussed in connection with FIG. 4a. Item 100
illustrates a collection of features extracted for different media
data items as indicated at 45 in FIG. 1. Specifically, line 100a
illustrates the metadata for five different features F1 to F5 for
the first audio file indicated at 45a in FIG. 1. Analogously, item
100b illustrates the extracted features for the second media data
item of block 45 in FIG. 1.
[0054] The set of media data items, i.e., the collection for which
a certain profile is calculated, is represented by the five
exemplary media data items in FIG. 4a at 100. It is visible from
FIG. 4a that all five features are quantitative features, which are
given in a certain value range, which extends between 0 and 10. In
the embodiment of FIG. 4a, each quantitative feature has the same
value range extending between 0 and 10. In other embodiments,
however, the value range for a feature can include negative values
as well. Furthermore, the value ranges for different features can
be different from each other so that, for example, a first features
has a value range between 0 and 100 and a second feature has a
value range between -10 and +10, and so on.
[0055] For calculating the raw music DNA, i.e., a raw collection
profile for the collection 100, the corresponding features F1 for
all items are added and an average is calculated. In the exemplary
case, where an arithmetic average has been calculated, this
operation incurs additions of the individual features F1 and a
subsequent division by the number of participants. The result for
the first feature is 4.4. Analogously, the averages for the
corresponding features F2 to F5 are calculated, and they are, in
this embodiment, 4.2, 3.0, 0.2 and 8.2. In one embodiment, the
vector of these averages, i.e., the vectors (4.4; 4.2; 3.0; 0.2;
8.2) is the collection profile or music DNA, when the media data
items are audio files.
[0056] The components of the collection profile correspond to the
individual features F1 to F5. Therefore, the collection profile has
five components, since five features have contributed to the
collection profile. Depending on certain applications, the number
of components of the collection profile can be smaller. In this
situation, individual components could be averaged so that, for
example, the collection profile only has three components, wherein
the first component corresponds to an average between the first
feature F1 and the second feature F2, the second component
corresponds to an average of the third feature F3 and the fourth
feature F4, and the third component would correspond to the fifth
feature F5.
[0057] Furthermore, it becomes clear from the example in FIG. 4a
that each DNA component has the same number range as the underlying
feature. Since, however, DNA components represent averages between
features, the number range of a DNA component could also be smaller
than the number range of an individual feature.
[0058] In the example of FIG. 4a, where the number of components of
the music DNA is identical to the number of individual features
used to extract the music DNA, a distance between the music DNA and
a vector of features can easily be calculated as indicated at 110.
In the embodiment, the distance measure D corresponds to the
geometric distance between a certain music DNA and a certain
feature vector. Since the feature vectors, i.e., the metadata for a
media data item, are quantitative vectors, and since the music DNA
is a quantitative vector as well, a quantitative distance measure D
can be calculated between each feature vector and a music DNA. An
exemplary equation is given at 110, but there can also be applied
other distance measures, such as non-quadratic distances, higher
order distances, logarithmic distances, etc.
[0059] A matching or non-matching decision can be taken when the
distance measure D, which is exemplarily calculated at 110, is
smaller than a predefined distance. This predefined distance can be
set by the user and determines the number of matching hits for a
certain search. Additionally, one can also determine a match
between the music DNA and a media data item having a feature vector
which results in the smallest distance D among all other feature
vectors in the set.
[0060] The music DNA is specifically useful, since it has a similar
appearance as a feature vector. Therefore, the music DNA can be
easily compared to a plurality of media data items to find a
matching media data item and, alternatively or additionally, the
music DNA can be compared to a plurality of different music DNAs.
Thus, the matching procedure, when a music DNA is matched with
other music DNAs, is identical to the matching procedure, when a
music DNA is matched with feature vectors for media data items.
[0061] Subsequently, reference is made to column 112, in which the
distance measure D between the music DNA of the user and all music
items in the user's collection 100 is indicated. When, for example,
the predefined distance would be set to 1.0, then media data items
1, 2, 3 would be located as matching items. When, however, the
selection would be done in that only the best matching media file
will be extracted, then the media data item having the ID number 2
would be selected as the result of the database matching
operation.
[0062] As discussed in connection with step 84 or 86 in FIG. 3, a
usage behavior or a user taste can be used to influence the values,
i.e., the components of the music DNA. One way for introducing the
usage behavior would be to weight the feature vectors 100a, 100b
before averaging to calculate the raw music DNA in order to
increase the weight of a recently played or often played piece
compared to a non-recently played or seldom played piece.
[0063] Alternatively, the different components of the calculated
music DNA can be weighted in order to account for a user taste.
When, for example, one feature, such as feature F2, represents the
percussiveness of an audio piece, and when the user wishes to have
pieces having a high percussiveness but, regarding the other
features, matching with his music DNA, the user could implement his
specific music taste in several ways. One way would be to
synthetically increase the percussiveness in each feature vector so
that, in the end, the features F2 of the different media data items
have a higher value due to the additional weighting/refinement.
[0064] Alternatively, the user could increase the percussiveness
average, i.e., the value 4.2 at 100 in FIG. 4a to, for example,
8.4. Then, automatically, the distance measure calculation will
result in hits having a higher percussiveness. When, for example,
the user wishes to have audio pieces selected, which have a
percussiveness close to 4.2, the user could influence the distance
measure calculation D at 110 in order to weight the second term by
a weighting factor higher than 1. Then, the second feature is
favored with respect to the other features in the distance
calculation and a result of a database matching operation will
favor pieces having a good matching second features F2.
[0065] In order to favor certain moods of music, several features
can be collectively influenced by certain weighting factors. To
this end, a table is of advantage in which a certain mood is mapped
to certain weighting factors for specific features. Then, the
discussed modifications of the feature vectors or the music DNA or
the distance calculation can be performed using not only a single
factor as discussed, but a collection of advantageously different
factors.
[0066] FIG. 4b illustrates the result list of a matching operation
which has been performed by using, for example, the matching rule
110 with or without influencing the DNA components or with or
without having certain weighting factors in the distance
calculation. When a matching operation is performed in the data
base 100, a distance is obtained for each media item, and,
typically, some media items will match well, and other media items
will not match well. Thus, the result list of a matching operation
can be given as a sorted list where the media item having the
smallest distance will be the first media item and where the media
item having the second to smallest distance will be number 2
etc.
[0067] FIG. 4c illustrates a plot of the distance versus the number
in the sorted result list, and it becomes clear either from the
result list in table representation of FIG. 4b or from the result
list in the plot representation in FIG. 4c that there is a certain
gap between media item 5 and media item 6 with respect to the
distance. Specifically, media items between No. 1 and No. 5 have
distances between 0.6 and 2.2, which are quite close to each other,
while the distance of media item No. 6 is much higher than the
distance of media No. 5.
[0068] In accordance with the embodiment, a confidence threshold is
determined. Generally, for determining this confidence threshold,
the change of the distance from one media item in the sorted list
to the next media item in the sorted list is determined and, as
soon as this change or difference between two adjacent distances is
larger than a threshold, a confidence threshold is determined as
illustrated in FIG. 4c.
[0069] Then, in accordance with this embodiment, a confidence
measure is mapped to all media items which have a distance below
the confidence threshold, and the confidence measure for all media
items, which have a distance above the threshold, is set to 0. This
results in a mapping as indicated in the last column of FIG. 4b,
and the mapping rule for mapping the confidence measures in this
example is indicated as a linear mapping rule.
[0070] In other implementations, however, there can also be a
compressed or a logarithmic or generally non-linear mapping rule
for mapping the confidence measures to media items. Furthermore,
the determination of the confidence threshold can be performed
based on different criteria such as a mix between a distance
increase between two adjacent media items and the number of media
items having a distance below the distance, in which the distance
increase occurs. Other criteria can include a mix between a
distance increase, a number of media items below a suggested
threshold and an absolute value of the distance. For example, it
can be preset that all media items having a distance below a
distance threshold of two receive a confidence measure above 0%.
Generally, the confidence measure operation is to free the user
from any statistically irrelevant data which are obtained by using
the quantitative collection profile. In other words, the user is
not interested in any distance measures or in any "hits" in the
sorted result list, which have such a high distance that it can be
said for sure that the user is not interested in such a media item
in the current search. On the other hand, mapping a confidence
measure between 0 and 100% will provide the user with a tool which
is familiar to the user and which allows the user to quickly
determine the usefulness of the results of a matching
operation.
[0071] FIG. 4d illustrates a flow chart of the procedure which can
be implemented as a method or as an apparatus in order to produce
the examples of FIGS. 4a, 4b, 4c in a general form.
[0072] In step 420, the matching operation comprises the
calculation of a distance D(i) for each media item in the
collection.
[0073] In a subsequent step 421, the result list is sorted so that
D(i) increases from low to high as exemplarily indicated in FIG.
4b.
[0074] In a step 422, a confidence threshold TH is determined.
Furthermore, the distance D(TH) of the confidence threshold TH is
determined. In the FIG. 4c embodiment, the confidence threshold TH
is equal to 5 which is the current number of the media item, which
is equal to or below the confidence threshold. The distance at this
threshold is equal to 2.2 in the FIG. 4c embodiment.
[0075] In a step 423, a confidence measure is mapped to each media
item having a distance below the threshold distance. The linear
mapping rule is indicated at 419 in FIG. 4c in order to obtain the
confidence measure value for each item i having a number in the
result list which is below or equal to a threshold.
[0076] In a step 424, the final result list which only includes the
media items having a number below or equal to the threshold, where
each media item has an associated confidence measure is used.
Different usage scenarios for this final result list exist. In a
first implementation, the result list can be stored for later use.
In a different usage scenario, the result list can be played as it
is so that, in a play list, number 1 is played. Subsequently number
2 is played, subsequently number 3 is played etc. This will be done
until the first media item having a confidence measure of 0% is
played, which is media item Number 5 in the FIG. 4b embodiment. An
alternative usage scenario is to perform a random play operation
within this final result list so that only media items of this
final result list are played, but in a random order which is not
influenced by the confidence measure order. A further usage
scenario is to use this final result list for any kind of play list
generation different from a random play list.
[0077] FIG. 6 illustrates an exemplary user DNA, which is a
sophisticated user profile which may automatically be generated
from the user's music collection. Specifically, the user DNA in
FIG. 6 comprises qualitative features 120a, which, in this
embodiment, indicate the genre of the audio pieces and,
additionally, comprises quantitative features 120b. Therefore, in
an embodiment, the quantitative features 102d are used and,
additionally, qualitative features 120a are added. Depending on the
implementation, specific maximum values 122 can be excluded from a
matching operation or can be included. Therefore, a matching
operation in a database can be performed without the energy strong
feature and the color bright feature or, depending on the user
instruction, these extreme values can be used as well.
[0078] This feature makes sure that, on the one hand, extraction
errors are minimized and, on the other hand, a selection is not
completely dominated by a certain dominant feature unless the user
wishes to include this dominant feature to a database
operation.
[0079] Subsequently, advantageous metadata implementations are
discussed.
[0080] MPEG-7 is providing XML formatted information about content,
but in MPEG-7, this data is not directly linked or embedded into
the audio content itself. The MP7 file format is closing this gap
by combining and embedding both--audio content and describing
metadata in order to obtain MPEG-7 enriched files.
[0081] FIG. 8 illustrates an embedding of MPEG-7 data into an MP3
file, an MP4 file (AAC file), a WMA file, an AC3 file, etc.
[0082] To this end, an audio content portion 190 is provided as
usual and a metadata portion 200 is added. This metadata portion
200 can be an ID3 V2 portion which has a public MPEG-7 part and a
premium MPEG-7 part 200a, 200b. The public portion 200a includes a
selection of basic tags for free access, such as genre, music
color, speech music discrimination. The premium MPEG-7 portion 200b
includes additional tags for premium users, such as information on
a subgenre and features for tempo/BPM (beats per minute),
percussiveness, vocal detection, aggressiveness, key, mood,
segmentation, solo instrument, etc. These content-related features
have been pre-calculated and entered into the second portion 200b.
Advantageously, this second portion 200b is encrypted so that only
a premium user, i.e., a user having a certain right to use this
portion or having paid for the usage of this portion, can use this
portion while a non-premium user can only use the first portion
200a. The public portion, however, includes at least describing
metadata, such as artist, title, album, etc.
[0083] Generally, a file format in accordance with the present
invention comprises an audio recording portion 190, which may be an
MPEG audio portion, such as MP3, AAC or WMA. The metadata portion
may be in the ID3 V2 (version 2) format, which includes standard
metadata, such as artist, title, album, etc.
[0084] The public metadata portion 200a is in accordance with
MPEG-7 and includes openly accessible metadata stored in MPEG-7 XML
for recommendation and advanced navigation, such as genre, music
color, speech music discrimination.
[0085] The additional metadata portion 200b includes additional
metadata going beyond public MPEG-7 metadata. These metadata are
stored in public-key encrypted format and can only be accessed by
premium subscribers and operators or distributors who have acquired
a decryption key.
[0086] Depending on the level of access rights, advanced tags, such
as subgenre, tempo/BPM, percussiveness, vocal detection,
aggressiveness, key, mood, segmentation, solo instrument, etc., and
P2P (peer-to-peer) license information are provided, or features
which are also enabling tracing of works.
[0087] FIG. 7 illustrates a feature overview including qualitative
and quantitative features. Qualitative features are genre,
subgenre, speech/music discrimination, ID3 data. Quantitative
features are tempo/beat determination, percussiveness, music color,
vocal detection, aggressiveness, segmentation, key, mood, solo
instrument and soundslike fingerprint.
[0088] The public MPEG-7 tags indicated at 200a in FIG. 8 are, as
stated before, openly accessible metadata stored in MPEG-7 XML
format. This portion includes a selection of high-level tags
describing a characteristic of a song for advanced search, such as
genre, music color, speech music discrimination. This portion 200a
furthermore provides the base for a basic tag-based recommendation
and advanced navigation. Additionally, information in the portion
200a provide the base for a basic automated playlist
generation.
[0089] The information included in the premium/second portion 200b
includes advanced high-level tags and low-level audio fingerprint
information, such as melody, subgenre, tempo/BPM, percussiveness,
vocal detection, aggressiveness, key, mood segmentation, solo
instrument, etc.
[0090] The information in this portion allows a more exact search
and navigation than the information in the first portion 200a.
Furthermore, the information in portion 200b provides a song
segmentation and segment-based search and comparison. It also
allows a base for exact recommendation and intelligent playlist
generation. Advantageously, the information in portion 200b is
stored in a public-key-encrypted format and, thus, can only be
accessed by premium subscribers.
[0091] Furthermore, different levels of access restrictions can be
provided, such as for premium consumers, rights holders, operators,
etc. The portion 200b furthermore contains license and sales
information, and this portion furthermore enables a peer-to-peer
distribution and tracing of works. The sophisticated high-level
tags in the second portion 200b can furthermore be used to find
cover versions and plagiarism.
[0092] Generally, MPEG-7 is a multimedia content description
interface, which is a standard for describing the multimedia
content data that supports some degree of interpretation of the
information meaning. MPEG-1 is for storage and retrieval of video
and audio on storage media. MPEG-2 is for storage and retrieval in
digital television. MPEG-4 codes content as objects and enables
those objects to be manipulated individually or collectively.
MPEG-7 is a standard for describing features of a content, and
MPEG-21 is for digital rights management.
[0093] Advantageously, the extraction and generation of an MPEG-7
enriched song is performed by the following main steps.
[0094] In a first step, the audio material is decoded and the
available ID3 information is imported.
[0095] In a second step, the audio material is split into frequency
bands and short-time frames.
[0096] In a third step, the calculation of low-level features for
each frame, such as modulation spectrum, noise likeness, etc., is
performed.
[0097] In a fourth step, a soundslike fingerprint for low-level
recommendation is generated.
[0098] In a further step, the song segmentation is calculated via
self-similarity information to distinguish chorus, verse, intro and
solos.
[0099] In a further step, high-level features are calculated based
on low-level features either for the whole song or for each
segment.
[0100] Then, in a final step, the extracted data is embedded into
the audio file in order to create an MPEG-7 enriched song.
Specifically, the new feature information not already available in
the first portion 200a is inserted into portion 200b of FIG. 8.
[0101] Generally, an apparatus for generating a collection of
different media data items in accordance with an embodiment of the
present invention is illustrated in FIG. 10. The apparatus
comprises a feature extractor 300 for extracting, for a plurality
of media data items, at least two different features described in a
content of a media data item using the media data items.
Specifically, the feature extractor can be implemented in order to
accurately calculate these features, such as the high-level
features as discussed before. Due to the flexibility and variety of
these features, the features need a variable amount of bits to be
stored in the metadata portion 200. In order to provide flexibility
and future adaptability, a metadata generator 310 is provided. The
metadata generator generates a metadata portion 200 for each media
data item of the plurality of media data items, where the metadata
generator is adapted to generate a metadata header 320 as
illustrated in FIG. 11, to generate an information on a content
characteristic represented by each feature, and to generate a
metadata payload portion 330 associated with the metadata header,
the metadata payload portion having the at least two different
features for the media data item. In the FIG. 11 embodiment, the
header 320 therefore comprises the length information for the
complete payload portion or comprises, for each feature separately,
a separate length information. Other ways to indicate where the
information for a certain feature ends and the information for a
further feature starts are useful in addition, such as the
implementation of synchronization marks, etc.
[0102] Furthermore, in order to allow future extension capabilities
for adding additional features, information on the meaning of each
feature, i.e., feature F1 is, for example, a mood feature and F2
is, for example, an aggressiveness feature, etc., is included as
well. This information can be included in a standardized order or
can be explicitly included in the metadata header 320.
[0103] When FIG. 11 and FIG. 8 are compared, it is to be noted that
metadata header 320 and the metadata payload portion 330 can,
together, constitute the second portion 200b. Alternatively, the
metadata header 320 can be a header relating to the first portion
200a as well as to the second portion 200b. In this case, the
second portion 200b would include the metadata header 320, and the
information included in the first portion 200a of FIG. 8, i.e., the
public tags, would form an additional metadata payload portion
which might precede the metadata payload portion 330 or which might
follow the metadata payload portion 330 in a certain format.
[0104] Advantageously, the metadata header furthermore comprises
information on additional fields of the metadata payload portion,
which are not illustrated in FIG. 11, i.e., access information,
license and sales information or any additional information related
to the way how the media data file associated with the metadata
information can be used, distributed, etc. In this scenario, the
metadata header 320 would include additional length information and
description information of such additional metadata payload portion
fields as well.
[0105] In an embodiment, the feature extractor is operative to
detect a newly stored media data item in the collection. Then, the
profile creator 20 is operative to retrieve an earlier collected
profile and to manipulate the earlier collected profile based on
the features of the newly stored media data item to create an
updated collection profile. This feature allows to reduce the
amount of time and resources for fully calculating a new profile
when the user has added a single new item only. In an embodiment,
the calculation of the profile takes place in the background, i.e.,
when the user does not use the device and the device is running in
an idle state. This makes sure that as soon as the user wishes to
use her or his device, the user can enjoy the full computing
capabilities of the device, since the profile calculation is
stopped in this case.
[0106] Subsequently, reference will be made to FIG. 2 and FIGS. 5a
to 5c in order to illustrate a plurality of ways of how the
inventive collection profile can be advantageously used for
performing a communication between two entities. Generally, as
illustrated in FIG. 2, the inventive apparatus for communicating
information from a first entity to a second entity comprises the
collection profile information generator 70, which actually
calculates the collection profile from a collection of metadata
items or which receives a readily calculated profile information
via a wireless or a cable interface. In the first possibility, in
which the collection profile information generator actually
calculates the collection profile, the collection profile is from
the first entity, when the collection profile information generator
is located in the first entity. In the alternative, in which a
readily calculated collection profile is received by item 70 via a
wireless or cable interface, the collection profile refers to a
collection of media data items within a first entity, wherein item
70 is located in the second entity.
[0107] The information handler 60 uses the collection profile in
order to find a media data item having a matching feature vector or
to find a user or a user group having a matching collection
profile. In the former case, a collection profile is compared to a
plurality of different feature vectors for media data items, where
in the second case, a collection profile obtained from item 70 is
compared to the plurality of collection profiles of other
entities.
[0108] Based on this matching operation, information is sent or
received, wherein this information not necessarily has to include
the matching result or the collection profile but can relate to any
product/service or any information related to a product/service
determined in response to the collection profile matching
operation.
[0109] In a certain embodiment, which is illustrated in FIG. 5a,
the collection profile information generator resides in a sound
processor/generator of an audio listening room, such as a bar, a
lobby or anything else, or provides streaming data for a certain
Internet user group. Therefore, the collection profile information
generator 70 represents a second entity and collects profiles from
users in a user group or a room, as indicated at 500. The users in
a group or a room represent the first entities.
[0110] In a step 502, the device at the second entity represented
by the information handler 60 of FIG. 2 calculates an average
profile among the profiles from the users received in step 50.
Then, based on this average profile, audio pieces are searched in a
database, which match with this average profile, as indicated at
step 504. Particularly, a database matching operation as discussed
in connection with FIG. 4a is performed by using the average
profile on the one hand and a collection of feature vectors or an
audio database on the other hand in order to find matching audio
pieces. Then, in a step 506, the selected audio pieces for the user
group or the room are played or streamed, when an Internet user
group scenario is considered.
[0111] FIG. 5b illustrates an alternative embodiment for an
advantageous usage of the collection profile, where all steps take
place in the first entity, i.e., in the user premises.
Specifically, the user determines his user profile by analyzing his
collection of media data items indicated at 510 in FIG. 5b. Based
on his calculated collection profile, the user finds a suitable
user group by matching different group profiles with his own
profile, as indicated at 512. In this embodiment, the result of the
matching operation in step 512 performed by the information handler
60 in FIG. 2 will only output the best matching result, since this
is what the user wants to have. In other words, the user wants to
join a user group where the other members of the user group have
the same or a very similar music taste.
[0112] Then, in a step 514, the information handler 60 of FIG. 2,
which already performed the operation 512, now sends a join request
message to the selected user group, as indicated in step 514.
Alternatively, however, the user could send his own user profile
determined in step 510 to a remote user group matching service.
Then, the determination of the user profile will be performed in
the first entity 510 and the user group matching service performed
in step 512 will be in the second entity, and the step 514 would be
performed in a different way, i.e., that the (with respect to the
user) remote second entity will send information on the matching
user group to the user.
[0113] A further alternative of an advantageous usage of the
collection profile is illustrated in FIG. 5c. The steps in FIG. 5c
will take place in an entity which is different from the entity
from which the user profile comes from. In step 520, a user profile
of a different entity is determined such as by wireless or cable
transmission. Then, a matching audio/video data item is searched
and found in the own database, as indicated in step 530. This step
is performed in the same entity in which the user profile of the
other entity is received. In a step 540, the matching item found in
step 530 is sent to the other entity. This mechanism consisting of
steps 520 to 540 may be used for a peer-to-peer distribution of
media data items, where the other entity can be sure to receive
only a metadata item which fulfils the taste of the other entity
due to the fact that the selection was done based on the user DNA
of the other entity.
[0114] An analogous procedure is illustrated by steps 550 and 560.
In step 550, a certain entity may use a user profile of a remote
entity for finding a matching product/service for such a user
profile. To this end, the information handler 60 of FIG. 2 will
perform a matching operation, which may be similar to the matching
operation regarding the distance in FIG. 4a, when each available
product/service has a certain feature vector. Alternatively, a user
profile may be input in a certain function, such as a table lookup
in order to find a best-matching product/service.
[0115] Then, in step 560, an advertisement for a matching product
or a matching service is delivered to the other entity by the
information handler 60. Thus, the advertisement pointing to a
certain product or service has been determined and forwarded to the
other entity based on the user DNA of the other entity.
[0116] In accordance with embodiments of the present invention, the
inventive collection profile enables new dimensions with community
building. The inventive user profile will enable various tailored
services, like self-pushing files in a peer-to-peer (super
distribution) network, a tailored background for visitors in music
bars/clubs and community Internet radio
(streaming/prodcasting).
[0117] Such Bach DNA communities can become interesting targets for
advertisers having products matching the profile of a Bach DNA, as
discussed with respect to items 550 and 560 of FIG. 5.
[0118] In the advantageous usage scenarios described above, the
music DMA has been used to influence/trigger/determine a
communication message from a first entity to a second entity, where
these two entities were users/persons/devices at different
geographic locations. FIG. 5d now illustrates an embodiment, in
which the user DNA is used by the user herself or himself for
generating her or his own play list or, generally stated, by
locating a media item which is to be played/rendered for her or his
own enjoyment.
[0119] To this end, the user retrieves the collection profile of
her or his own collection as indicated in step 581. In one
implementation, this collection profile is used in a matching
operation so that the user finds, from her or his own collection, a
media item which has the smallest distance. Alternatively, the user
can also determine, from her or his own collection, a final result
list as discussed in connection with FIG. 4c where the confidence
measures are indicated for this own collection. In this
implementation, the information handle 60 of FIG. 1 or the
information handle 60 of FIG. 2 will use a music DNA for matching
the data base of the user, and the result will be the play list
generation or the rendering of a media item from the user's own
collection for the enjoyment of the user, i.e., by the user device
in which the matching operation has been performed.
[0120] In the embodiment illustrated in FIG. 5d, however, the user
will modify her or his collection profile of her or his own
collection to obtain a modified profile as indicated in step 582.
The modification can be performed by the user so that the user can
set his current taste. When, for example, the user is in an
aggressive mood, then the user can modify her or his raw collection
profile so that a more aggressive profile is generated which,
however, nevertheless resembles the user's profile. In the end,
when a matching operation is performed using this modified profile
as indicated in step 583, the user will be in the position to enjoy
a media item which is, of course, more aggressive, but which
nevertheless conforms with the user's own music taste. Therefore,
the user does not just hear an aggressive audio track, for example,
but is in the position to hear a very specific audio track which,
on the one hand, is more aggressive and which, on the other hand,
however, nevertheless has characteristics which conform with the
user's general taste.
[0121] Other alternatives for modifying the collection profile in
step 582 are external situations. An external situation can, for
example, be a time of the day so that the music player of the user
calculates a different music DNA based on the raw collection
profile of step 581, for example for the morning or for example for
the evening of a day.
[0122] A different external situation can be the day in a week so
that, for example, a different modified profile is calculated for a
Sunday or a Saturday compared to, for example, a Tuesday or
Wednesday.
[0123] A further external situation can be a geographic location so
that the user's modified DNA is different when the user is, for
example, in Europe or in Asia or in the US.
[0124] Based on this modified profile, a matching operation in the
own media collection is performed as indicated in step 583, and the
result of this matching operation is used in the step 584 in order
to play the matching result for the user's own enjoyment or for
generating a play list in the user media device for playing certain
media items in a certain order.
[0125] The Bach DNA will be used in a communication scenario
connecting consumers with consumers, artists with artists, artists
with consumers in order to calculate/create unique communities.
This service can be based on a fee paid by the end user, the
Internet service provider or the advertiser for using the
advantageous Bach DNA tags included in section 200b of FIG. 8 for
the connecting consumers and artists.
[0126] Subscription models such as download and streaming, which
are based on a flat fee (compulsory or voluntary) or
advertisement-based services are allowing end users to access
unlimited music and tailor it for the user's personal DNA.
[0127] A super distribution extension to the subscription system
allowing the consumer to redistribute (share) music in a community
with other consumers and artists with similar music DNAs is made
possible by the inventive user DNA.
[0128] Bach DNA communities will become interesting targets for
advertisers having products matching the profile of a Bach DNA and
give more value per click. Such a situation is illustrated in FIG.
9, where, for certain moods, such as relax, exciting, happy, angry,
fear, sorrow, sad, certain products illustrated in FIG. 9 are
associated. The mood can be a certain explicit feature or can be
calculated by a combination of different features. In this case, a
rule for combining certain components of a collection profile
vector in a certain way results in a certain number, where this
number is then mapped to a certain mood. The association of a
certain product to a certain mood can be done via a table and the
selection of a certain advertisement as soon as a certain product
has been identified is straight forward.
[0129] Thus, the Bach user DNA together with the Bach DNA tags is
the key to perfectly personalized music recommendation.
[0130] Shop systems and communities can compare user DNA and thus
link to songs, artists and even other users with fitting music
taste.
[0131] Different to existing statistical solutions, such as "people
who have bought this also bought that . . . ", the DNA approach
provides a measure which really corresponds to the user and which
even works outside of the mainstream and allows up-to-date
suggestions.
[0132] The Bach user DNA can be applied to user groups to provide
several unique services, such as automated music recommendation or
self-pushing files in a peer-to-peer (super distribution) scenario.
Another service is the analysis of the music taste and of streams
within the community, and a further service is the automated
generation of live radio streams fitting that taste for a
(community) Internet radio and even for background music in bars or
clubs depending on their customers profiles, as discussed
before.
[0133] As already discussed in connection with FIG. 3, item 88, it
is a feature of the present invention to enrich a music DNA profile
of a sound collection stored in a customer's handset by additional
useful information. Such an additional information is, for example,
the brand of the handset and the model of the handset as well as
the mobile operator connecting the handset to a communication
system or to the inventive match-making service. Therefore, the
handset brand and model as well as the mobile operator may be
identified in real-time. In the new online music world, the handset
is the new carrier taking over from the CD. Therefore, the
characteristics of the handset and connecting operator may be
included in the customer music DNA profile. The advantage is to
discover any correlation between type of handset and type of music
profile. This provides an added value even more, when the MPEG-21
standard related to digital rights management becomes reality. Like
the ID3 V2 tags complement the inventive DNA with identification
title and producer/rights holders of a soundtrack, the handset
characteristics tags will identify the type of carrier and the name
of carrier manufacturer.
[0134] The music DNA/collection profile of the user allows exciting
application scenarios for a user which are not only related to
finding matching user groups, finding matching advertisements or
finding songs corresponding to the user taste and, additionally,
corresponding to other issues related to the user. The music DNA
additionally allows to open completely new scenarios when, for
example, two individuals swap their music DNAs. In this situation,
a first entity is in the position to experience the world of a
second different entity, which can result in astonishing
experiences for both users specifically when both users want to
learn more about each other. Therefore, a music DNA can be a key
feature in building and experiencing an alter ego situation.
[0135] The described embodiments are merely illustrative for the
principles of the present invention. It is understood that
modifications and variations of the arrangements and the details
described herein will be apparent to others skilled in the art. It
is the intent, therefore, to be limited only by the scope of the
impending patent claims and not by the specific details presented
by way of description and explanation of the embodiments
herein.
[0136] The inventive user collection profile can be stored on a
digital storage medium or a machine-readable carrier, in
particular, a disc, a DVD or a CD having electronically-readable
control signals stored thereon, which co-operate with programmable
computer systems such that the user profile is displayed or used
for database matching or used for a communication message.
[0137] Depending on certain implementation requirements of the
inventive methods, the inventive methods can be implemented in
hardware or in software. The implementation can be performed using
a digital storage medium, in particular, a disc, a DVD or a CD
having electronically-readable control signals stored thereon,
which co-operate with programmable computer systems such that the
inventive methods are performed. Generally, the present invention
is therefore a computer program product with a program code stored
on a machine-readable carrier, the program code being operated for
performing the inventive methods when the computer program product
runs on a computer. In other words, the inventive methods are,
therefore, a computer program having a program code for performing
at least one of the inventive methods when the computer program
runs on a computer.
[0138] While this invention has been described in terms of several
embodiments, there are alterations, permutations, and equivalents
which fall within the scope of this invention. It should also be
noted that there are many alternative ways of implementing the
methods and compositions of the present invention. It is therefore
intended that the following appended claims be interpreted as
including all such alterations, permutations, and equivalents as
fall within the true spirit and scope of the present invention.
* * * * *