U.S. patent application number 13/725624 was filed with the patent office on 2013-08-15 for state detection.
This patent application is currently assigned to Vodafone IP Licensing Limited. The applicant listed for this patent is Vodafone IP Licensing Limited. Invention is credited to Thomas Robert LOVETT, Eamonn Joseph O'NEILL, David POLLINGTON.
Application Number | 20130210480 13/725624 |
Document ID | / |
Family ID | 45572949 |
Filed Date | 2013-08-15 |
United States Patent
Application |
20130210480 |
Kind Code |
A1 |
POLLINGTON; David ; et
al. |
August 15, 2013 |
STATE DETECTION
Abstract
Method of determining a state of a mobile device, the mobile
device having a one or more sensors, the method comprising the
steps of: obtaining sensor data from the one or more sensors of the
mobile device in an initial state. Determining if the initial state
matches a previous state by comparing the obtained sensor data to
previous sensor data acquired with the mobile device in the
previous state. A mobile device includes logic to carry out this
method.
Inventors: |
POLLINGTON; David; (Newbury,
GB) ; LOVETT; Thomas Robert; (Newbury, GB) ;
O'NEILL; Eamonn Joseph; (Newbury, GB) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Vodafone IP Licensing Limited; |
|
|
US |
|
|
Assignee: |
Vodafone IP Licensing
Limited
Newbury
GB
|
Family ID: |
45572949 |
Appl. No.: |
13/725624 |
Filed: |
December 21, 2012 |
Current U.S.
Class: |
455/550.1 |
Current CPC
Class: |
H04M 2250/12 20130101;
H04W 8/22 20130101; H04M 1/72569 20130101; H04W 88/02 20130101 |
Class at
Publication: |
455/550.1 |
International
Class: |
H04W 8/22 20060101
H04W008/22 |
Foreign Application Data
Date |
Code |
Application Number |
Dec 22, 2011 |
GB |
1122206.4 |
Mar 23, 2012 |
GB |
1205160.3 |
Claims
1. A method of determining a state of a mobile device, the mobile
device having one or more sensors, the method comprising the steps
of: obtaining sensor data from the one or more sensors of the
mobile device in an initial state; and determining if the initial
state matches a previous state by comparing the obtained sensor
data to previous sensor data acquired with the mobile device in the
previous state, wherein the comparison of the obtained sensor data
to the previous sensor data is carried out on vector
representations of both data.
2. The method of claim 1, wherein the comparison uses, statistical
inference, frequentist inference, Bayesian inference, or
Hotelling's T.sup.2.
3. The method according to claim 1, wherein the previous sensor
data is retrieved from a database of sensor data acquired with the
mobile device in different states.
4. The method of claim 3 further comprising the step of storing the
obtained sensor data of the mobile device in the first state in the
database when no matches are found.
5. The method according to claim 1, wherein the previous sensor
data is stored in a database as vector representations.
6. The method according to claim 1 further comprising the step of
selecting a previous state to compare with the initial state based
on a nearest-neighbour search of the sensor data.
7. The method according to claim 1 further comprising the step of
selecting a previous state to compare with the initial state by
determining the most commonly matched previous states.
8. The method according to claim 1, wherein the sensors of the
mobile device are one or more selected from the group consisting
of: an accelerometer, a digital compass, a GPS receiver, a
microphone, an ambient light sensor, a thermometer, a radio
receiver, a proximity sensor, a Bluetooth receiver and a Wi-Fi
receiver.
9. The method according to claim 1 further comprising the step of
receiving a state identifier for the initial state and/or the
previous states.
10. A mobile device comprising: one or more sensors; and logic
configured to: obtain sensor data from the one or more sensors of
the mobile device in an initial state; determine if the initial
state matches a previous state by comparing the obtained sensor
data to previous sensor data acquired with the mobile device in the
previous state, wherein the comparison of the obtained sensor data
to the previous sensor data is carried out on vector
representations of both data.
11. The mobile device of claim 10, wherein the comparison uses,
statistical inference, frequentist inference, Bayesian inference,
or Hotelling's T.sup.2.
12. The mobile device according to claim 10 further comprising a
database configured to store sensor data acquired with the mobile
device in different states.
13. A computer program comprising program instructions that, when
executed on a computer cause the computer to perform the method of
claim 1.
14. A computer-readable medium carrying a computer program
according to claim 13.
15. A computer programmed to perform the method of claim 1.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority to United Kingdom
Application Number 1122206.4, filed on Dec. 21, 2011, and United
Kingdom Application Number 1205160.3, filed Mar. 23, 2012 the
entireties of which are incorporated herein by reference.
FIELD OF THE INVENTION
[0002] The present invention relates to a mobile device and method
of determining a state of the mobile device.
BACKGROUND OF THE INVENTION
[0003] Mobile devices and especially cell phones, may be configured
and operated in different ways depending on their particular state
or context. For example, the display may be dimmed when the mobile
device is in a state of low ambient illumination and the volume of
the ringer may be raised when the mobile device is in a state of
high ambient noise. A user may provide an input to the mobile
device when the state or context of the mobile device changes, in
order to actively alter operating characteristics or settings. For
example, if a user enters a meeting, they may wish to place their
mobile device in "meeting" mode, which disables the ringer, enables
the vibration alert and perhaps disables or dims the display.
[0004] There may be many other different states and that the mobile
device may operate within each requiring different settings or
operational responses from the mobile device. There are other
optimisations or functionality that may be used provided a
particular state is known. However, it is impractical and
inconvenient for the user to manually select or identify a
particular state whenever such a new state is encountered.
[0005] Therefore, there is required an improved mobile device and
method that may overcome these problems and disadvantages.
SUMMARY OF THE INVENTION
[0006] Against this background and in accordance with a first
aspect there is provided a method of determining a state of a
mobile device, the mobile device having one or more sensors, the
method comprising the steps of:
[0007] obtaining sensor data from the one or more sensors of the
mobile device in an initial state; and
[0008] determining if the initial state matches a previous state by
comparing the obtained sensor data to previous sensor data acquired
with the mobile device in the previous state. Therefore, the state
or context of the mobile device may be detected or determined
automatically. Furthermore, this avoids privacy issues as the state
may be acquired or determined within the device without necessarily
relying on or providing external information. This method may be
preferably carried out within a mobile device and so network
connections or external storage of personal data may be avoided.
Preferably, data from more than one sensor may be acquired and
compared simultaneously to provide more robust and accurate state
detection. The initial state may be similarly described as the
present or current state of the mobile device or the state at the
time the sensor data are obtained.
[0009] The comparison may be made by comparing patterns within or
of the obtained data against patterns in the previous data.
[0010] Advantageously, the comparison may use statistical
inference. This allows a match to be made even where sensor data
are not identical.
[0011] Advantageously, the comparison of the obtained sensor data
to the previous sensor data may be carried out on vector
representations of both data. The vector representations may be
codebook vectors. The obtained data may be converted to a vector
representation. The previous sensor data may also be in the form of
vector representation and preferably stored in the format to aid
comparison and matching and also to reduce storage space and
processing power requirements.
[0012] Optionally, the comparison may use statistical inference,
frequentist inference, Bayesian inference, or Hotelling's T.sup.2
distribution.
[0013] Preferably, the previous sensor data may be retrieved from a
database of sensor data acquired with the mobile device in
different states. Therefore, the sensor data may be stored locally
without requiring data transfer resources.
[0014] Optionally, the method may further comprise the step of
storing the obtained sensor data of the mobile device in the first
state in the database when no matches are found. This provides a
mechanism to update and increase the available matchable and
comparable data.
[0015] Preferably, the previous sensor data is stored in a database
as vector representations and preferably codebook vectors. Other
storage formats may be used. Vectors representations may be
compared more easily using statistical inference.
[0016] Optionally, the method may further comprise the step of
selecting a previous state to compare with the initial state based
on a nearest-neighbour search of the sensor data. This reduces the
processing and time necessary to make a match.
[0017] Optionally, the method may further comprise the step of
selecting a previous state to compare with the initial state by
determining the most commonly matched previous states. This reduces
the processing and time necessary to make a match as more commonly
matched results may be more likely to be matched in future.
[0018] Optionally, the sensors of the mobile device may be one or
more selected from the group consisting of: an accelerometer, a
digital compass, a GPS receiver, a microphone, an ambient light
sensor, a thermometer, a radio receiver, a proximity sensor, a
Bluetooth receiver and a Wi-Fi receiver.
[0019] Optionally, the method may further comprise the step of
receiving a state identifier for the initial state and/or the
previous states. This may allow easier configuration of known
states. This may involve semantically tagging the initial (and/or
previous) states. The state identifier may be received from a user.
User intervention may take the form of presenting the user with
matched or unmatched states (i.e. a set of states that were initial
states or previous states). The user may then provide a label,
name, identification or context as the state identifier for these
states if known to them. For example, the time and date may be
presented along with each state and the user can provide a
narrative or label for future use as a state identifier (e.g.
should a match be made against a previously identified state).
Unknown states may be presented for identification. Furthermore,
previously identified or similar states may be presented for
confirmation, updating or deletion if necessary.
[0020] Such state identifiers may be received from external sources
(in addition or instead of from the user). These external sources
may include social networks or other user data (e.g. crowd
sourcing).
[0021] According to a second aspect there is provided a mobile
device comprising:
[0022] one or more sensors; and
[0023] logic configured to: [0024] obtain sensor data from the one
or more sensors of the mobile device in an initial state; [0025]
determine if the initial state matches a previous state by
comparing the obtained sensor data to previous sensor data acquired
with the mobile device in the previous state. The logic may be
contained within a processor or processors of the mobile
device.
[0026] Advantageously, the comparison of the obtained sensor data
to the previous sensor data may be carried out on vector
representations of both data.
[0027] Optionally, the comparison uses statistical inference,
frequentist inference, Bayesian inference, or Hotelling's T.sup.2
distribution.
[0028] Optionally, the mobile device may further comprise a
database configured to store sensor data acquired with the mobile
device in different states. The database may be configured to store
fingerprints or vector representations of the sensor data instead
of or as well as the actual data.
[0029] Optionally, the method or logic of the mobile device may
further comprise:
[0030] obtaining further sensor data from the one or more sensors;
and
[0031] determining whether there has been a transition from the
initial state to a second state, based on a difference between the
sensor data obtained in the initial state and the further sensor
data. In other words, the method or logic within the mobile device
may further determine a state transition. Transition detection may
occur before, after or at the same time as state determination.
[0032] Optionally, the method or logic may be further configured to
obtain additional sensor data from a plurality of the additional
sensors if there has been a transition from the initial state to
the second state.
[0033] Optionally, the method or logic may be further configured to
determine there has not been a transition from the initial state to
a second state if the difference between the obtained sensor data
and obtained further sensor data does not exceed a predetermined
threshold.
[0034] The methods described above may be implemented as a computer
program comprising program instructions to operate a computer. The
computer program may be stored on a computer-readable medium or
sent as a signal.
[0035] It should be noted that any feature described above may be
used with any particular aspect or embodiment of the invention.
BRIEF DESCRIPTION OF THE FIGURES
[0036] The present invention may be put into practice in a number
of ways and embodiments will now be described by way of example
only and with reference to the accompanying drawings, in which:
[0037] FIG. 1 shows a schematic diagram of a mobile device in
accordance with an embodiment of the present invention;
[0038] FIG. 2 is a flow diagram illustrating the steps taken by a
mobile device for determining a context state transition and
identifying a context state;
[0039] FIG. 3 is a schematic diagram illustrating an example of
user context labelling, in accordance with an embodiment of the
present invention; and
[0040] FIG. 4 is a schematic diagram illustrating an example of
user context correction, in accordance with an embodiment of the
present invention.
[0041] It should be noted that the figures are illustrated for
simplicity and are not necessarily drawn to scale.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0042] Whilst various embodiments of the present invention are
described below, the invention is not limited to these embodiments
and variations of these embodiments may well fall within the scope
of the invention which is to be limited only by the appended
claims.
[0043] Further embodiments and specific examples may be found in
the appended Annex, "Event-based Context Sampling and Online
Interference with Mobile Devices", which is incorporated into this
description.
[0044] FIG. 1 illustrates a mobile device in accordance with an
embodiment of the present invention. Mobile device 10 comprises six
sensors, 12a-12f, logic 14 stored in a suitable medium of the
mobile device, such as a memory of the mobile device, and processor
16. Mobile device 10 is typically any portable device, such as a
mobile telephone, a laptop computer, a personal digital assistance,
etc. In other embodiments, the mobile device need not be easily
portable, and, for example, could be a desktop computer.
[0045] Sensors 12a-12f are configured to accept input from the
outside environment. For example, sensor 12a may be an ambient
light sensor and may comprise a photosensitive detector adapted to
measure a light level outside of mobile device 10. Sensor 12b may
be an accelerometer and may be configured to measure an
acceleration of mobile device 10 along a plurality of axes. Sensor
12c may be a GPS receiver and may be configured to measure a
geographic position of mobile device 10, by receiving relevant data
from GPS satellites. Sensors 12c-12f could comprise a network
sensor, such as a radio transceiver adapted to receive
communications from, and transmit them to, a mobile network of
device 10.
[0046] Logic 14 is operatively connected to each of sensors 12a-12f
and is configured to read a corresponding data output from each of
sensors 12a-12f. Processor 16 is connected to logic 14 and is
configured to read logic 14 so as to manipulate and process the
data outputs read from sensors 12a-12f. Mobile device 10 may
further comprise a memory (not shown) which may be configured to
store data outputs read from sensors 12a-12f. Processor 16 may then
be further configured to access data outputs stored in the memory,
and manipulate these stored values, and may combine them with fresh
data outputs read from one or more of sensors 12a-12f.
[0047] A database 19 may be used to store sensor data or
representations of sensor data.
[0048] Mobile device 10 may determine its current state. In what
follows, various advantages and particular implementations of the
present invention are described.
[0049] The activity of a user may be considered as one of a number
of discrete states--for instance, sitting at a desk, making a cup
of coffee in the kitchen, having lunch in the work canteen, etc. A
mobile device according to the present invention allows one to
identify characteristics of a particular state as it is encountered
and thereafter may determine the identity or category of the new
state that has been entered. A process of `sensor fusion` may be
used to derive a state that the individual (and their handset) has
entered. By building up heuristic knowledge of the states that the
user commonly enters, the mobile device can create a profile of the
user and their most common (or habitual) states, and thereby use
this for improving state recognition and/or predicting future
states based on states that have already been encountered within a
given time period.
[0050] For example, rather than simply using an accelerometer to
determine movement, the mobile device may use algorithms across
multiple sensor-based data sets to characterise the context of the
individual. For instance, whilst in a meeting, an individual may
leave their handset on the table, in a pocket, or be `fiddling`
with it--each of these scenarios may be creating very different
sensor profiles, but all pertain to the same static state of `in a
meeting`.
[0051] 1. Context State Definition
[0052] The mobile device captures data from one or more sensors
(accelerometer, digital compass, GPS receiver, Wi-Fi receiver,
proximity sensor, Bluetooth receiver, Cell ID, microphone, ambient
light sensor, etc.) and uses `sensor fusion` techniques and
algorithms that are processed on the handset to create a unique
fingerprint for the current context of an individual/handset (note
that this fingerprint is not a static snapshot, but rather a
characterisation created by sensor outputs within a specific
temporal pattern and amplitude, and, when combined with the outputs
of additional sensors, uniquely identify a state associated with an
individual/handset). `Sensor fusion` is the term used to describe
multiple sensors generating respective data outputs, these outputs
then being read by logic of the mobile device so as to determine a
particular state.
[0053] States may be defined as the characterisation of a context
of an individual in such a way as to have semantic meaning, either
to the individual or across a group of individuals. A `common`
state most likely refers to a context that is linked to a place
that is frequently visited (such as `home` or `work` on a macro
level, or else `desk` or `canteen` on a micro level). An `ad hoc`
state represents a transient context which is common across a group
of individuals at a time and place (such as being stuck in a
traffic jam on a motorway). When an `adhoc` state is repeated
consistently (e.g., a traffic jam at a known traffic black spot),
then such a state may then become a `common` state with a specific
semantic tag (e.g. traffic jam on Heathrow Airport spur road). The
respective outputs from multiple sensors of the mobile device may
be used to characterise and assign an `output signature` to each of
these states, or contexts.
[0054] 2. Context State Transition Triggering
[0055] Detecting the current state may be triggered by various
conditions or simply achieved at regular and/or predefined
intervals. In one embodiment, the mobile device may identify when a
sensor output exceeds given thresholds, hence signalling a change
in user state, possibly from a known state to an unknown state. The
mobile device may be effectively moving from a static state to a
dynamic one, and vice versa--e.g. getting up from your desk at work
and walking to somewhere else. Upon transiting from one state to
another (e.g. from a static state to a dynamic state), the mobile
device may activate and analyse the outputs from additional sensors
(sensor fusion) in order to identify what the new state is (e.g. as
described under "Context state definition"). The sensors that are
employed may depend on the type of state the user is transitioning
into, e.g. location-based sensors for a static state, and
activity-based sensors for a dynamic state.
[0056] 3. Context State Semantic Labelling & Correction
[0057] In order for the state information to be useful across a
variety of different applications, it may be necessary to attach a
semantic meaning to the state (at the very least, to those states
that are common/habitual for the individual in question, or,
alternatively, to all states, in which case generic tags may be
used to describe ad hoc states that are common across multiple
individuals, such as being in a traffic jam). There are multiple
mechanisms that could be employed to attribute semantic meaning to
each static state: [0058] States may be populated into an
`inference` database and presented to the user for manual tagging
and/or correction of inferred tags--to minimise the load on the
user, this could be constrained to only common states. [0059] If
the user uses a social media check-in service, then this semantic
tag may also be used to tag that particular state recognised by the
system (by leveraging public social media APIs to obtain details of
the activity that the individual in question posted on the social
media site). [0060] Similarly, if the individual uses #tags to
represent their current context, this information could also be
harvested by open Twitter APIs for semantic tagging. [0061] If the
individual redeems a voucher (via their mobile handset) at a Point
of Sale, then the information relating to which store the
individual is in could also be used for semantically tagging the
state (context) of the individual. [0062] If the user enters a
state that is recognised as similar to an historic state recorded
for another individual, then there may be the option of applying
that state's tag to the user's state (effectively, a form of
implicit crowd-sourcing of tags). Grouping state patterns across
multiple users may generate a library of semantically tagged
patterns against which the pattern/state of a new user can be
compared, thereby determining the place of the user without the
need for explicit training for that user.
[0063] Advantages of the present invention include: [0064] The
invention may provide rich, real-time contextual information for a
given user that can be used within a wide range of services. [0065]
The invention may be carried out on a mobile device without the
need for user interaction (although user interaction may improve
performance, such as in the case of embodiments directed to
labelling and correction). [0066] By recognising and historically
creating a record of user states, the invention may provide
context/state information to a 3rd party system without explicit
user interaction (as would be the case with current social media
systems that require the user to check-in manually each time a
location is visited). In effect, a user would only have to
`check-in` (i.e. semantically tag) a location (state) once, and
thereafter the system could automatically `check-in` the user
whenever that place was visited. [0067] Furthermore, the invention
may (but not necessarily does) operate entirely on the handset
without sending user data to 3rd parties, hence protecting the
privacy of the user (i.e. contextual information may be derived on
the handset and provided solely to applications on the handset to
which the user has granted authorisation to use that
information)--in doing so, the user is given a greater degree of
privacy and control over the use of the data.
[0068] The present invention may be used in various different
applications, examples of which are described below.
[0069] Transport (Traffic Congestion)
[0070] By defining journey paths as a succession of states
(waypoints), common states (waypoints) across a multitude of users
may indicate traffic blackspots (either historically or at a
particular time). Note that, in this example, it may be assumed
that a state may only be triggered when the user/car is deemed to
have stopped (e.g. transitioned from a dynamic, moving state to a
static, stationary state). As well as identifying traffic
congestion, the invention could also be used to determine which
motorway service stations are most popular for people on a journey
travelling between two given points.
[0071] Shopping Mall
[0072] The invention may provide an effective solution for
determining which store a user has visited, or is currently in, by
identifying a unique state for each store (e.g., a combination of
data outputs relating to location, movement, ambient lighting and
ambient noise, determined through the mobile device's sensors).
Taking a semantic approach to identifying which stores a user has
visited obviates the need to deploy dedicated infrastructure for
providing indoor positioning.
[0073] Network Optimisation
[0074] The invention may compare user state (context) against
behaviour to identify where entering a given state might lead to a
common behaviour (service usage). By linking behaviour with state,
and predicting states across multiple users, a mobile network
operator may then predict likely demand (usage) of services, and
hence optimise the mobile network accordingly. This may include
noting which content is downloaded when users enter a given state
(which in this case may be defined according to the data outputs
read from location sensor and a time sensor), and precaching
content accordingly, based on the number of users likely to be
entering that state (place) in a given timeframe.
[0075] Network Offload
[0076] Increasingly, mobile network operators are looking at
various network offload strategies that leverage small cell
technologies (e.g. Wi-Fi and Femto cells) to better manage and
optimise the capacity and service offered by a cellular network.
With the present invention, it may be possible for handsets to
search for WiFi points at the moment of state transition rather
than periodically, hence increasing the likelihood of a WiFi
network being joined as soon as the user arrives at a new
destination, and increasing battery efficiency. This is in contrast
to handsets continually polling for available WiFi networks.
[0077] Enterprise Presence
[0078] Awareness of a state or changes of state could be used in an
enterprise context for providing more accurate presence
information. Typically, personal computer systems monitor keyboard
activity as an indicator of whether an individual is available or
`away`, but this can deliver false information if, for instance,
the individual is reading at their desk, or is engaged in a
conference call, rather than actively using the PC keyboard. With
the present invention, using the mobile handset as a trigger point
in addition to the keyboard, the resulting enriched presence system
would have a better understanding of the individual's actual
context. By building up a time-stamped record of the states that a
user enters, it may be possible to start predicting which state an
individual might be transitioning to either on a short-term basis
(e.g. `has left desk, predict going to office canteen`) or a
longer-term basis (e.g. populating the user's calendar with
availability information for the use of other individuals--e.g.
`David tends to work from home on a Friday so don't book meetings
at the office location, but a conference call would be
acceptable`).
[0079] User Profiling
[0080] By building up a record of the states (places) that are
common to an individual, it becomes possible to assign semantic
inference to that individual. For instance, a high proportion of
states tagged as relating to shopping may be a leading indicator
that the individual in question likes to shop (or works in a
shop).
[0081] FIG. 2 illustrates the steps taken in a method according to
a preferred embodiment of the present invention. In this
embodiment, a mobile device (e.g. mobile device 10 of the
embodiment described above) determines that a state transition has
occurred, and in particular may determine when the mobile device
has transitioned from a known state to an unknown state. The mobile
device may then identify the unknown state, thereby rendering it a
known state.
[0082] In step 21, an internal clock in the mobile device advances
one timestep forwards. An arbitrary time unit may be used, such as
ten seconds. Thus, the mobile device may perform method 20 every
timestep, for example every ten seconds.
[0083] In step 22, the mobile device takes a motion reading from a
motion sensor (e.g. an accelerometer). The received data is a
vector describing component-wise intensity of motion. If the mobile
device is stationary, for example, then the motion sensor reads a
motion vector of zero intensity for all three components. If the
mobile device is moving at a constant velocity (constant magnitude
and direction), then the motion sensor may read the corresponding
intensity of motion.
[0084] In step 23, the mobile device classifies whether the device
is in a `static` or a `dynamic` state, given the current motion
vector. The mobile device may use a low-pass filter in series with
a pre-trained binary classifier, together with a moving average
filter (which stores a short history of previous binary classifier
outputs). Further details of this step are described in the Annex
especially with reference to FIG. 2 of the Annex.
[0085] In step 24, the mobile device determines whether the output
state type at this timestep is equivalent in {0,1} to the previous
timestep's output state type. If yes, no transition is classified,
and the next iteration may begin. If no, a transition is
classified, triggering the inference process and beginning the next
timestep iteration.
[0086] Determining the state of a mobile device starts at step 25.
The new context state type (static or dynamic) may have been
determined at step 23. If so determined, the mobile device 10 may
obtain, acquire or take a set of data samples from that particular
type's sensors or from any or all of the sensors. The mobile device
10 may then derive a fingerprint or vector representation of the
obtained sensor data. A determination may be made as to whether the
present or initial state of the mobile device 10 matches a previous
state or a state that has previously been observed or encountered.
If no match is made then the initial state may be determined to be
a new, previously unvisited state. The determination or match may
be carried out on vector representations of the data preferably
using statistical inference including frequentist and Bayesian or
Hotelling's T.sup.2 distribution.
[0087] Example inference techniques are described in the Annex
under "Inference and Classification" on page 10.
[0088] If the state is a previously unvisited state, then the
mobile device may assign a different label to the classified or
initial state (step 26). If the state is a previously visited
state, the mobile device may assign the label of the identified
previously visited state to the newly classified state (step 27).
This state may now be an instance of the previously visited state
and may be stored in database 19, preferably within the mobile
device 10. In order to classify a new state, the mobile device may
assign a new, unique label to the newly classified state. This
state is now the first instance of the newly labelled state. In
step 28, the inference process terminates.
[0089] FIG. 3 illustrates a method of determining a state of a
mobile device, in accordance with a preferred embodiment of the
present invention. A state may be described by a set of codebook
vectors or fingerprints of the sensor data. Once a state has been
identified, the user may assign a human-readable label to the
state. Once assigned, the label may be displayed to the user on
future classifications of the state (e.g. in FIG. 3, state #1234).
Before this label is applied, the state may be identified only by a
unique ID assigned by the identification process. Through an
interface, the user may label the state and, upon labelling, all
instances of that state (both past and future) may be associated
with this label unless re-labelled.
[0090] FIG. 4 illustrates a method of user context creation. The
user may correct the system identification which, in turn, may
affect future system outputs. There are three types of corrective
action:
[0091] Add: The user creates a state that partially overwrites (in
the database or other storage facility within or external to the
mobile device 10) one or more identified states. The states that
are partially overwritten by this new state may be used to create
the codebook vectors for the new state, and the user's new label is
assigned to it. Furthermore, it is assigned a unique ID by the
system.
[0092] Edit: There are two types of edit for an identified state:
i) the user re-labelling the identified state as a previously
labelled state (Case 1); and ii) the user re-labelling the
identified state as a new, previously unidentified state (Case
2).
[0093] Delete: The user removes an identified state. If there is
only one instance of this state, it is removed entirely. If not,
this instance alone is removed and previous updates to the codebook
vectors are undone.
[0094] In FIG. 4, at step 42, the state with the label `Desk`
exists, and so, upon user correction, the sensor data from the
incorrect instance of `Meeting` may be used to update the codebook
vectors for the state `Desk`. At step 44, the state with the label
`Cafe` does not exist, and so a new state is created with a unique
ID and label `Cafe`. The sensor data from the incorrect instance of
`Meeting` are used to create the codebook vectors for the state
`Cafe`.
[0095] As will be appreciated by the skilled person, details of the
above embodiment may be varied without departing from the scope of
the present invention, as defined by the appended claims.
[0096] For example, the database may be external to the mobile
device, with data being transferred across a mobile network. The
previous data may have been acquired from a different mobile device
or may be synthetic data.
[0097] Many combinations, modifications, or alterations to the
features of the above embodiments will be readily apparent to the
skilled person and are intended to form part of the invention. Any
of the features described specifically relating to one embodiment
or example may be used in any other embodiment by making the
appropriate changes.
* * * * *