U.S. patent application number 10/526190 was filed with the patent office on 2006-10-12 for method and apparatus for envelope detection and enhancement of pitch cue of audio signals.
Invention is credited to Peter John Blamey, Hugh McDermott, James F. Patrick, Brett Anthony Swanson.
Application Number | 20060227986 10/526190 |
Document ID | / |
Family ID | 31979111 |
Filed Date | 2006-10-12 |
United States Patent
Application |
20060227986 |
Kind Code |
A1 |
Swanson; Brett Anthony ; et
al. |
October 12, 2006 |
Method and apparatus for envelope detection and enhancement of
pitch cue of audio signals
Abstract
A method and apparatus for detecting an envelope of an audio
signal, and a method and apparatus for enhancing the pitch cue of
an audio signal perceived by a cochlear implant patient where the
audio signal is processed and input to an implant device of the
recipient. The methods and apparatuses use techniques such as
filtering, rectifying, detecting peak values, sampling, resetting,
comparing and multiplying various signals to detect the envelope or
enhance the pitch cue of the audio signal.
Inventors: |
Swanson; Brett Anthony;
(Meadowbank, AU) ; Blamey; Peter John; (Mt.
Waverly, AU) ; McDermott; Hugh; (Mt. Macedon, AU)
; Patrick; James F.; (Roseville, AU) |
Correspondence
Address: |
JAGTIANI + GUTTAG
10363-A DEMOCRACY LANE
FAIRFAX
VA
22030
US
|
Family ID: |
31979111 |
Appl. No.: |
10/526190 |
Filed: |
September 2, 2003 |
PCT Filed: |
September 2, 2003 |
PCT NO: |
PCT/AU03/01130 |
371 Date: |
June 15, 2006 |
Current U.S.
Class: |
381/312 |
Current CPC
Class: |
G11C 27/02 20130101;
A61N 1/36038 20170801; H04R 25/606 20130101; G11C 27/024
20130101 |
Class at
Publication: |
381/312 |
International
Class: |
H04R 25/00 20060101
H04R025/00 |
Foreign Application Data
Date |
Code |
Application Number |
Sep 2, 2002 |
AU |
2002951165 |
Aug 18, 2003 |
AU |
2003904405 |
Claims
1-74. (canceled)
75. A method of detecting an envelope of an audio signal comprising
the steps of: filtering the audio signal to produce a filtered
audio signal; rectifying the filtered audio signal to produce a
rectified signal; detecting peak values of the rectified signal to
produce a detected signal; sampling the detected signal at
predetermined time intervals to produce samples; and resetting the
detected signal immediately after sampling.
76. The method according to claim 75, wherein the rectifying step
uses half wave rectification.
77. The method according to claim 75, wherein the rectifying step
uses full wave rectification.
78. The method according to claim 75, wherein the detected peak
values remain at a substantially constant value prior to the
sampling step.
79. The method according to claim 78, wherein the detected signal
or detected signals is reset substantially to zero.
80. The method according to claim 75, wherein the sampling rate
used in the sampling step is relatively low compared to frequency
components in the filtered audio signal.
81. The method according to claim 75, wherein the audio signal is
input to a cochlear implant device.
82. An apparatus for detecting an envelope of an audio signal
comprising: means for filtering the audio signal to produce a
filtered audio signal; means for rectifying the filtered audio
signal to produce a rectified signal; means for detecting the peak
values of the rectified signal to produce a detected signal; means
for sampling the detected signal at predetermined time intervals to
produce samples; and means for resetting the means for detecting
immediately after sampling, such that the detected signal is reset
immediately following sampling.
83. The apparatus according to claim 82, wherein the means for
rectifying is one or more full wave rectifiers.
84. The apparatus according to claim 82, wherein the means for
rectifying is one or more half wave rectifiers.
85. The apparatus according to claim 82, wherein the detected peak
values remain at a substantially constant value prior to
sampling.
86. The apparatus according to claim 85, wherein the detected
signal or detected signals is reset substantially to zero.
87. The apparatus according to claim 82, wherein the sampling rate
used by the means for sampling is relatively low compared to
frequency components in the filtered audio signal.
88. The apparatus according to claim 82, wherein the audio signal
is input to a cochlear implant device.
Description
FIELD OF THE INVENTION
[0001] The present invention relates to hearing prosthesis and to
sound processing devices and methods associated with hearing
prosthesis. In particular, the present invention relates to an
apparatus and method of envelope detection that is simple to
implement in both analog circuitry or digital signal processing and
assists cochlear implant recipients to better perceive changes in
the amplitude of speech than is currently the case. Furthermore,
the invention relates to an apparatus and method for enhancing the
pitch cue of an audio signal perceived by a cochlear implant
recipient.
BACKGROUND OF THE INVENTION
[0002] In many people who are profoundly deaf, the reason for
deafness is absence of, or destruction of, the hair cells in the
cochlea which transduce acoustic signals into nerve impulses. These
people are unable to derive suitable benefit from conventional
hearing aid systems, no matter how loud the acoustic stimulus is
made, because there is damage to or absence of the mechanism for
nerve impulses to be generated from sound in the normal manner.
[0003] It is for this purpose that cochlear implant systems have
been developed. Such systems bypass the hair cells in the cochlea
and directly deliver electrical stimulation to the auditory nerve
fibres, thereby allowing the brain to perceive a hearing sensation
resembling the natural hearing sensation normally delivered to the
auditory nerve. U.S. Pat. No. 4,532,930, also in the name of the
applicant and the contents of which are incorporated herein by
reference, provides a description of one type of traditional
cochlear implant system.
[0004] Typically, cochlear implant systems have consisted of
essentially two components, an external component commonly referred
to as a processor unit and an internal implanted component commonly
referred to as a receiver/stimulator unit. Traditionally, both of
these components have cooperated together to provide the sound
sensation to a user. The external component has traditionally
consisted of a microphone for detecting sounds, such as speech and
environmental sounds, a speech processor that converts the detected
sounds, particularly speech, into a coded signal, a power source
such as a battery, and an external transmitter coil.
[0005] The coded signal output by the speech processor is
transmitted transcutaneously to the implanted stimulator/receiver
unit situated within a recess of the temporal bone of the user.
This transcutaneous transmission occurs via the external
transmitter coil which is positioned to communicate with an
implanted receiver coil provided with the stimulator/receiver
unit.
[0006] This communication serves two essential purposes, firstly to
transcutaneously transmit the coded sound signal and secondly to
provide power to the implanted stimulator/receiver unit.
Conventionally, this link has been in the form of a radio frequency
(RF) link, but other such links have been proposed and implemented
with varying degrees of success.
[0007] The implanted stimulator/receiver unit traditionally
includes a receiver coil that receives the coded signal and power
from the external processor component, and a stimulator that
processes the coded signal and outputs a stimulation signal to an
intracochlear electrode assembly which applies the electrical
stimulation directly to the auditory nerve producing a hearing
sensation corresponding to the original detected sound.
[0008] Traditionally, the external componentry has been carried on
the body of the user, such as in a pocket of the user's clothing, a
belt pouch or in a harness, while the microphone has been mounted
on a clip behind the ear or on the lapel of the user.
[0009] More recently, due in the main to improvements in
technology, the physical dimensions of the speech processor have
been able to be reduced allowing for the external componentry to be
housed in a small unit capable of being worn behind the ear of the
user. This unit allows the microphone, power unit and the speech
processor to be housed in a single unit capable of being discretely
worn behind the ear, with the external transmitter coil still
positioned on the side of the user's head to allow for the
transmission of the coded sound signal from the speech processor
and power to the implanted stimulator unit. It is envisaged that
with further technological advancements the system components will
be able to be fully implanted within the head of the recipient,
providing a totally invisible device.
[0010] As the ability to perceive sound is of fundamental
importance to cochlear implant recipients, the ability to reproduce
sound and the percepts of speech via electrical stimulation using a
cochlear prosthesis is one of the major challenges of this
technology. It is the speech processor that provides the link
between the acoustic representation of speech and the pattern of
neural discharges which the stimulator of the implant is able to
induce, and which the recipient experiences as hearing sensations.
Many speech-processing strategies such as Continuous Inter-leaved
Sampling (CIS), and those based on spectral maxima SPEAK and ACE,
have been proposed to improve the quality of the sensation as
perceived by the recipient, in a number of different sound
environments.
[0011] These strategies utilise envelope detection for processing
the output of a series of filters, however a disadvantage of such
systems is that the output of the envelope detector typically
includes a large amount of ripple and/or the desired envelope
becomes excessively smeared out. This has the disadvantage of
adversely affecting the temporal cues that are important in
consonant perception. Other implementations of the strategies may
result in the output having a ripple which is aliased causing the
stimulation amplitude to vary with a frequency which is not present
in the input sound. Such a ripple can modulate at a beat frequency
which can give erroneous pitch cues to the implant recipient.
[0012] Another prior system called quadrature envelope detection,
although producing an envelope which substantially contains no
ripple and is not smeared out, has the disadvantage that it is
complex and requires twice as many band pass filters as there are
frequency channels which results in additional cost and complexity.
Furthermore, the need to provide the function of squaring and
square root operations is generally not practical in analogue
circuitry.
[0013] In normal hearing, the inner hair cells only response to
movement of the basilar membrane in one direction. They tend to
fire in phase with the basilar membrane response. This is known as
"phase-locking"; it preserves the timing content of the basilar
membrane response, and it is believed to be important for pitch
perception. At high stimulation rates, the fine timing content
generally has not been taken into account and therefore implant
recipients have not been able to adequately perceive pitch in an
audio signal. The present invention substantially preserves the
fine timing content of the band-pass filter outputs, and provides
an additional pitch cue to the cochlear implant recipient. It
requires high stimulation rates.
[0014] The present invention is therefore related to improving the
manner in which an audio signal is processed so that the quality of
sound reproduced via the electrical stimulation is substantially
maintained.
SUMMARY OF THE INVENTION
[0015] According to a first aspect of the invention, there is
provided a method of detecting an envelope of an audio signal
comprising the steps of: [0016] filtering said audio signal to
produce a filtered audio signal; [0017] rectifying the filtered
audio signal to produce a rectified signal; [0018] detecting the
peak values of the rectified signal to produce a detected signal;
[0019] sampling the detected signal at predetermined time intervals
to produce samples; and [0020] resetting the detected signal
immediately after sampling.
[0021] According to a second aspect of the invention there is
provided a method of detecting an envelope of an audio signal
comprising the steps, of: [0022] filtering the audio signal into
multiple filtered audio signals; [0023] rectifying each of the
mutiple filtered audio signals into respective multiple rectified
signals; [0024] detecting peak values of each of the multiple
rectified signals to produce detected signals; [0025] sampling each
of the detected signals at predetermined time intervals to produce
samples; and [0026] resetting each of the detected signals
immediately after sampling.
[0027] Preferably, the rectifying step involves using either half
wave rectification (HWR) or full wave rectification (FWR).
Preferably, each of the detected peak values remain at a
substantially constant value prior to sampling. Preferably, after
each sample the method further comprises the step of resetting the
detected signal or detected signals, and more particularly
resetting the detecting signal or detected signals substantially to
zero.
[0028] The sampling rate may be relatively low compared to the
frequency components in the filtered audio signal. Preferably the
audio signal is input to a cochlear implant device. According to a
third aspect of the invention, there is provided apparatus for
detecting an envelope of an audio signal comprising: [0029] means
for filtering the audio signal to produce a filtered audio signal;
[0030] means for rectifying the filtered audio signal to produce a
rectified signal; [0031] means for detecting the peak values of the
rectified signal to produce a detected signal; [0032] means for
sampling the detected signal at predetermined time intervals to
produce samples; and [0033] means for resetting the means for
detecting immediately after sampling, such that the detected signal
is reset immediately following sampling.
[0034] According to a fourth aspect of the invention there is
provided apparatus for detecting an envelope of an audio signal
comprising: [0035] means for filtering the audio signal into
multiple filtered audio signals; [0036] means for rectifying each
of the multiple filtered audio signals into respective multiple
rectified signals; [0037] means for detecting the peak values of
each of the multiple rectified signals to produce detected signals;
[0038] means for sampling each of the detected signals at
predetermined time intervals to produce samples; and [0039] means
for resetting the means for detecting immediately after sampling,
such that each of the detected signals are reset immediately
following sampling.
[0040] The present invention through processing the filtered signal
in one or more frequency ranges of interest, provides an improved
method of estimating the amount of energy present in a frequency
band used by a cochlear implant.
[0041] According to a fifth aspect of the invention there is
provided a method of enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, wherein the audio signal
is processed and input to an implant device of the recipient, the
method comprising the steps of: [0042] filtering the audio signal
to produce a filtered audio signal; [0043] half-wave rectifying the
filtered audio signal to produce a half-wave rectified signal; and
[0044] sampling the half-wave rectified signal at predetermined
time intervals.
[0045] This is in contrast with the prior art which aims to produce
a smooth envelope signal which varies slowly compared to the centre
frequency of the filter, thereby removing all of the fin timing
structure of the filtered signal.
[0046] According to a sixth aspect of the invention there is
provided a method of enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, wherein the audio signal
is processed and input to an implant device of the recipient, the
method comprising the steps of: [0047] filtering the audio signal
to produce a filtered audio signal; [0048] envelope detecting the
filtered audio signal to produce an envelope detected signal;
[0049] comparing the filtered audio signal to produce a gating
signal having one of two Boolean states; [0050] multiplying the
gating signal with the envelope detected signal to produce a
multiplied signal; and [0051] sampling the multiplied signal at
predetermined time intervals.
[0052] The step of envelope detection may involve quadrature
envelope detection such that the envelope detected signal is
produced using In-phase and quadrature phase filtered components of
the audio signal. The gating signal may be produced from an
in-phase filtered component of the audio signal. The filtering step
may involve using quadrature filters such that the audio signal is
filtered into in-phase and quadrature-phase components.
[0053] For sampling at high frequencies, the method may further
comprise the step of detecting the peaks of the multiplied signal
and resetting the multiplied signal, then sampling the multiplied
signal at predetermined time intervals.
[0054] According to a seventh aspect of the invention there is
provided a method of enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, wherein the audio signal
is processed and input to an implant device of the recipient, the
method comprising the steps of: [0055] filtering the audio signal
to produce a filtered audio signal; [0056] envelope detecting the
filtered audio signal to produce an envelope detected signal;
[0057] comparing the filtered audio signal to produce a gating
signal having one of two Boolean states; [0058] multiplying the
gating signal with the envelope detected signal to produce a
multiplied signal; [0059] detecting the peak values of and
resetting the multiplied signal to produce a peak detected and
reset multiplied signal.
[0060] The step of filtering may involve using quadrature filters
such that the audio signal is filtered into in-phase and
quadrature-phase components. The step of envelope detection may be
quadrature envelope detection such that the envelope detected
signal is based on the in-phase and quadrature-phase filtered
components of the audio signal.
[0061] Thus at high frequencies, preferably more than four times
the audio frequency, the method conveys fine timing content of the
filter output signal giving an enhanced pitch cue to the implant
recipient. Whilst at lower frequencies, preferably less than twice
the audio frequency, the method implements envelope detection.
[0062] The method may comprise applying the peak detected and reset
multiplied signal to apical electrode channels and to basal
electrode channels.
[0063] The method may comprise the step of sampling at relatively
high frequencies (high stimulation rates) for use by apical
electrode channels to obtain enhanced pitch cues, corresponding to
responses to low frequency signals.
[0064] The method may further comprise the step of sampling at
relatively low frequencies (low stimulation rates) using envelope
detection applied to basal electrode channels, corresponding to
responses to high frequency signals.
[0065] According to an eighth aspect of the invention, there is
provided an apparatus for enhancing the pitch cue of an audio
signal perceived by a cochlear implant recipient, the audio signal
being processed and input to an implant device of the recipient,
the apparatus comprising: [0066] means for filtering the audio
signal to produce a filtered audio signal; [0067] means for
half-wave rectifying the filtered audio signal to produce a
half-wave rectified signal; and [0068] means for sampling the
half-wave rectified signal at predetermined time intervals.
[0069] According to a ninth aspect of the invention, there is
provided apparatus for enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, the audio signal being
processed and input to an implant device of the recipient, the
apparatus comprising: [0070] means for filtering the audio signal
to produce the filtered audio signal; [0071] means for envelope
detecting the filtered audio signal to produce an envelope detected
signal; [0072] comparator means for producing a gating signal
having one of two Boolean states; [0073] means for multiplying the
gating signal with the envelope detected signal to produce a
multiplied signal; and [0074] means for sampling the multiplied
signal at predetermined time intervals. A
[0075] According to a tenth aspect of the invention, there is
provided apparatus for enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, the audio signal being
processed and input to an implant device of the recipient, the
apparatus comprising: [0076] means for filtering the audio signal
to produce a filtered audio signal; [0077] means for envelope
detecting the filtered audio signal to produce an envelope detected
signal; [0078] comparator means for producing a gating signal
having one of two Boolean states; [0079] means for multiplying the
gating signal with the envelope detected signal to produce a
multiplied signal; and [0080] means for detecting the peak values
of and resetting the multiplied signal to produce a peak detected
and reset multiplied signal.
[0081] The envelope detection means may include quadrature envelope
detection means. The filter means may include in-phase filter means
and quadrature-phase filter means. The apparatus may further
include circuit means for producing the envelope detected signal
based on values of the outputs to the in-phase and quadrature-phase
filter means. The comparator means may have at its input, the
output from the in-phase filter means.
[0082] According to an eleventh aspect of the invention there is
provided a method of enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, wherein the audio signal
is processed and input to an implant device of the recipient, the
method comprising the steps of: [0083] filtering the audio signal
to produce a filtered audio signal; and [0084] sampling the
filtered audio signal to produce samples; [0085] wherein the
samples are synchronized with the filtered audio signal.
[0086] Preferably the sampling step involves using a clock
synchroniser comprising bursts of pulses separated by a fixed time
interval with the leading pulse in each burst of pulses being
synchronised to the phase of the filtered audio signal. Preferably
the leading pulse occurs at a fixed time interval after the rising
zero crossing of the filtered audio signal, such that only positive
cycles of the audio wave form are sampled.
[0087] According to a twelfth aspect of the invention there is
provided a method of enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, wherein the audio signal
is processed and input to an implant device of the recipient, the
method comprising the steps of: [0088] filtering the audio signal
to produce a filtered audio signal; [0089] envelope detecting the
filtered audio signal to produce an envelope detected signal; and
[0090] sampling the envelope detected signal; [0091] wherein the
samples of the envelope detected signal are synchronised with the
filtered audio signal.
[0092] Preferably the sampling step involves using a clock
synchroniser that generates bursts of clock pulses separated by a
fixed time interval with the leading pulse in each burst of pulses
being synchronised to the phase of a portion of the filtered audio
signal. The step of envelope detection may involve quadrature
envelope detection with the leading pulse in each burst of pulses
of the clock synchroniser being synchronised to the in-phase
filtered signal. Preferably the leading pulse occurs at a fixed
time interval after the rising zero crossing of the filtered audio
signal, such that only positive cycles of the audio wave form are
sampled.
[0093] According to a thirteenth aspect of the invention there is
provided an apparatus for enhancing the pitch cue of an audio
signal perceived by a cochlear implant recipient, the audio signal
being processed and input to an implant device of the recipient,
the apparatus comprising: [0094] means for filtering the audio
signal to produce a filtered audio signal; and [0095] means for
sampling the filtered audio signal at predetermined time intervals
to produce samples; [0096] wherein the samples are synchronised
with the filtered audio signal.
[0097] According to a fourteenth aspect of the invention there is
provided apparatus for enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, the audio signal being
processed and input to an implant device of the recipient, the
apparatus comprising: [0098] means for filtering the audio signal
to produce a filtered audio signal; [0099] means for envelope
detecting the filtered audio signal to produce an envelope detected
signal; and [0100] means for sampling the envelope detected signal
to produce samples; [0101] wherein the samples of the envelope
detected signal are synchronised with the filtered audio
signal.
[0102] According to a fifteenth aspect of the invention there is
provided, in a multiple channel cochlear implant system permitting
sequential stimulation, a method of enhancing the pitch cue of an
audio signal perceived by a cochlear implant recipient, wherein the
audio signal is processed and input to an implant device of the
recipient, the method comprising the steps of: [0103] filtering the
audio signal to produce a filtered audio signal; [0104] sampling
the filtered audio signal to produce samples; and [0105]
synchronising the samples of the filtered audio signal using a
selection means and a series of master clock pulses, such that on
each master clock pulse no more than one channel is selected by the
selection means.
[0106] Preferably each channel has a low to high transition on a
channel enable signal. Each channel enable signal is preferably
input to the selection means and passed through the selection means
with controllable delay on each channel. Preferably where more than
one channel enable signal goes high on a single master clock pulse,
one channel is selected as previously described, with the remaining
channels delayed by successive master clock periods.
[0107] According to a sixteenth aspect of the invention there is
provided a method of enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, wherein the audio signal
is processed and input to an implant device of the recipient, the
method comprising the steps of: [0108] filtering the audio signal
to produce a filtered audio signal; [0109] half-wave rectifying the
filtered audio signal to produce a half-wave rectified signal;
[0110] detecting the peak values of the half-wave rectified signal
and resetting the detected peak values to produce a reset detected
signal; and [0111] sampling the reset detected signal at
predetermined time intervals.
[0112] According to a seventeenth aspect of the invention there is
provided apparatus for enhancing the pitch cue of an audio signal
perceived by a cochlear implant recipient, the audio signal being
processed and input to an implant device of the recipient, the
apparatus comprising: [0113] means for filtering the audio signal
to produce a filtered audio signal; [0114] means for half-wave
rectifying the filtered audio signal to produce a half-wave
rectified signal; [0115] means for detecting the peak values of the
half-wave rectified signal and resetting the detected peak values
to produce a reset detected signal; and [0116] means for sampling
the reset detected signal at predetermined time intervals.
[0117] It is to be understood that any of the embodiments described
in terms of acting on an audio signal to produce a filtered or
rectified signal can be extended to the case of filtering the audio
signal into multiple filtered audio signals and rectifying each of
the multiple filtered audio signals into respective multiple
rectified signals. Furthermore, a number of detected signals can be
sampled and reset or a number of rectified signals can be sampled.
Thus generally it is to be understood that multiple signals derived
from the original audio signal and acted upon, such as sampling,
rectifying, detecting, filtering, multiplying, comparing are
included as further embodiments.
BRIEF DESCRIPTION OF THE DRAWINGS
[0118] By way of example only, preferred embodiments of the
invention will now be described with reference to the accompanying
drawings, in which:
[0119] FIG. 1 is a pictorial representation of a conventional
cochlear implant system;
[0120] FIG. 2 is a block diagram showing the basic function of a
speech processor of a cochlear implant system;
[0121] FIG. 3 is a block diagram of the overall signal flow of a
conventional speech processing strategy;
[0122] FIG. 4 is a block diagram of a prior art envelope detection
method;
[0123] FIG. 5 is a graphical representation of the signal flow of
the prior art envelope detection method of FIG. 4;
[0124] FIG. 6 is a circuit diagram of the prior art envelope
detection method of FIG. 4;
[0125] FIG. 7 is a graphical representation of the signal flow of
the prior art envelope detection method of FIG. 4 with the
full-wave rectifier replaced with a half-wave rectifier;
[0126] FIG. 8 is a block diagram of an alternative prior art
envelope detection method;
[0127] FIG. 9 is a graphical representation of the signal flow of
the prior art envelope detection method of FIG. 8;
[0128] FIG. 10 is a block diagram of the envelope detection method
according to one aspect of the present invention;
[0129] FIG. 11 is a graphical representation of the signal flow of
one aspect of the present invention using full-wave
rectification;
[0130] FIG. 12 is a graphical representation of the signal flow of
one aspect of the present invention using half-wave
rectification;
[0131] FIG. 13 is a circuit showing one aspect of the present
invention using half wave rectification and
Peak-Detect-and-Reset;
[0132] FIG. 14 is a block diagram of an envelope detection method
according to a second aspect of the present invention;
[0133] FIG. 15 is a graphical representation of signals associated
with the circuit of FIG. 14;
[0134] FIG. 16 is a block diagram of a further scheme for envelope
detection but including a peak detect and reset stage;
[0135] FIG. 17 is a block diagram of a circuit according to a
further embodiment of the invention that enables enhanced pitch cue
of an audio signal and includes a half wave gating arrangement;
[0136] FIG. 18 is a block diagram similar to FIG. 17 wherein an
envelope detection circuit is represented as a quadratic envelope
detector;
[0137] FIG. 19 is a graphical representation of signals associated
with the circuit of FIG. 18;
[0138] FIG. 20 is a block diagram of a circuit arrangement similar
to FIG. 17 but including a peak detect and reset circuit;
[0139] FIG. 21 is a block diagram similar to FIG. 20 where the
envelope the detector circuit is represented as a quadrature
envelope detector;
[0140] FIG. 22 is a graphical representation of the signals
associated with the circuit of FIG. 21;
[0141] FIG. 23 is a graphical representation comparing the signals
associated with a uniform clock versus a synchronised clock;
[0142] FIG. 24 is a block diagram of a circuit arrangement of the
HWR-sync embodiment of the present invention;
[0143] FIG. 25 is a block diagram of a circuit arrangement of the
HWG-sync embodiment of the present invention; and
[0144] FIG. 26 is a clock synchroniser as used with the embodiments
described in relation to FIGS. 23 to 25.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS OF THE INVENTION
[0145] Before describing the features of the present invention, it
is appropriate to briefly describe the construction of one type of
known cochlear implant system with reference to FIG. 1.
[0146] Known cochlear implants 10 typically consist of two main
components, an external lo component including a speech processor
29, and an internal component including an implanted receiver and
stimulator unit 22. The external component includes a microphone
27. The speech processor 29 is, in this illustration, constructed
and arranged so that it can fit behind the outer ear 11 and is held
in place behind the outer ear 11 via an ear-hook arrangement (not
shown). Alternative versions may be worn on the body. Attached to
the speech processor 29 is a transmitter coil 24 that transmits
electrical signals to the implanted unit 22 via a radio frequency
(RF) link.
[0147] The implanted component includes a receiver coil 23 for
receiving power and data from the transmitter coil 24. A cable 21
extends from the implanted receiver and stimulator unit 22 to the
cochlea 12 and terminates in an electrode array 20. The signals
thus received are applied by the array 20 to the basilar membrane 8
and the nerve cells within the cochlea 12 thereby stimulating the
auditory nerve 9. The operation of such a device is described, for
example, in U.S. Pat. No. 4,532,930.
[0148] In order to appreciate the basic function of a speech
processor 30, such as that shown in FIG. 1, reference is made to
FIG. 2. As is shown, the speech processor 30 takes an audio signal
(usually from a microphone 32) and processes this signal according
to a particular speech processing strategy, to produce a signal
that contains stimulation information for an implant 34.
Conventionally, this signal is an RF signal that is
transcutaneously transmitted to the implant 34.
[0149] FIG. 3 represents an example of a typical block diagram of
speech processing in relation to cochlear implants. As is shown, a
microphone 36 detects an audio signal with this signal being
received by an analog front end 38. The analog front end 38, or
audio pre-processor, typically includes a preamplifier that
amplifies the very low signal from the microphone 36 to a level
which can be easily handled by the rest of the signal processing.
This analog front end 38 may also include user sensitivity signals
to assist in the pre-processing, such as gain controls and
sensitivity controls which can be set by the user to desirable
settings.
[0150] The next processing stage is a filter bank 40, which
typically consists of a set of band-pass filters that cover the
input frequency range. Each filter has a different centre frequency
allowing signals in one bands of frequencies to pass through whilst
other frequencies are blocked. The frequency bands may be based on
critical bands, for example these bands may be roughly linearly
spaced below 1000 Hz, and logarithmically spaced above 1000 Hz.
Each filter may be allocated to one channel (or pair of electrodes)
and as a result the number of filters may be equal to the number of
channels. The output of this processing stage is the envelopes of
the filtered signals, with the envelope being an estimate of the
instantaneous power in the corresponding spectral band. An envelope
detector processes the output of each filter to provide an estimate
of the amount of energy in the frequency band. By sampling the
envelope of each filter, the amplitude of the electrical
stimulation pulses can be controlled.
[0151] Following the filter bank stage 40 where a continuous set of
output signals are provided for each band-pass filter employed, the
Filter Bank outputs must then be sampled so that a sequence of
stimulation frames can be determined. This is performed by a sample
and selection unit 42. There are a number of strategies which
employ different sampling and selection techniques at this stage of
the signal processing, for example, CIS, SPEAK and ACE previously
referred to. As each filter is usually allocated to one channel,
the filter output sampling rate can be the same as the stimulation
rate on that channel.
[0152] In the CIS strategy, all filter bank output samples are
selected and the corresponding channels are then stimulated
sequentially. In the SPEAK and ACE strategies, a subset of channels
is selected for stimulation with the channels selected being those
that have the largest envelope amplitude at the sampling time.
[0153] The final stage of processing is the amplitude mapping
stage, referred to as the loudness growth function 44. In this
stage, the dynamic range of the envelope signal is compressed by a
loudness growth function 44 so that all sounds are mapped between
the threshold level (T) and the maximum comfort level (C) of the
channel to ensure that delivery of the stimulus is detected at the
appropriate sound intensity level by the user.
[0154] As is shown in FIG. 3, this signal can then be RF encoded
through encoder 46 and sent to the implanted stimulator for
delivery via the intracochlear electrodes.
[0155] As can be appreciated in the above summary of speech
processing strategies, at the heart of all the strategies is a
filterbank, which splits the sound into a number of frequency
bands. Therefore a fundamental aspect of such speech processing
strategies is the function of the envelope detector of each filter
which provides an estimate of the amount of energy present in each
frequency band. It is in this regard that the present invention
relates.
[0156] In the prior art, a number of different methods of detecting
the envelope of each filter have been proposed. One such prior art
method is shown in FIGS. 4 and 5, which utilises a rectifier and
peak detector.
[0157] In this method a band pass filter (BPF) 50 receives an audio
input signal. In FIG. 5 and each of the examples that follow, this
audio input signal is shown as a short burst of a 350 Hz pure tone
(uppermost signal), although it should be appreciated that this
input audio signal could be a sound signal as detected by a
microphone. In the example shown, the BPF has a centre frequency of
375 Hz and the output of the BPF is shown in FIG. 5.
[0158] In order to detect the envelope of this filter the signal is
passed through a rectifier 52, such as a full wave rectifier (FWR),
the output of which is also shown in FIG. 5. This signal is then
passed through a Peak Detector 54 that operates so that when its
input signal attempts to rise above its output signal, its output
signal follows its input signal; and when the input signal falls
below the output signal, the output signal gradually decays. In
order for this system to work satisfactorily, a compromise must be
made in choosing the decay time constant. In FIG. 5 a fast peak
detector output is shown that is obtained when the decay time is
short. The disadvantage here is that the output has a large amount
of ripple, as can be clearly seen in the signal. A slow peak
detector output is also shown that is obtained when the decay time
is long. This reduces the ripple, as is evident in the bottom
signal of FIG. 5, but it causes the desired envelope to be
excessively smeared out. This may adversely affect the temporal
cues that are important in consonant perception.
[0159] In order to derive the amplitude of the electrical
stimulation pulses associated with the audio signal, the output of
the peak detector is sampled by sampler 56. In a preferred
embodiment, the sample rate is equal to the channel stimulation
rate, and in this example the stimulation rate is 250 Hz. FIG. 5
shows the output samples for both the fast and slow peak detectors,
wherein each sample is shown by a vertical line. With full-wave
rectification, the ripple has twice the frequency of the BPF
output. Because the stimulation rate is lower than this, the ripple
is aliased and causes the stimulation amplitude to vary at a
frequency not present in the input sound. This modulation is an
artefact of the processing and may be confused with the actual
modulation of the envelope of a voiced speech sound thereby
distorting the sound perceived by the implant recipient.
[0160] Instead of a full-wave rectifier used in the prior art
system discussed above, this may be replaced by a half-wave
rectifier. Furthermore, the half-wave rectifier and peak detector
functions can be combined into one circuit. A simple embodiment of
this, widely used in AM radio receivers, is shown in FIG. 6. In
this example the decay time constant T is determined by the
relationship, T=RC. As can be seen clearly in FIG. 7, using a
half-wave rectifier instead of a full-wave rectifier increases the
amplitude of the ripple, and thus causes more distortion of the
envelope samples.
[0161] Both of the above mentioned prior art systems may also be
implemented digitally. In this case there are two sample rates: the
processing sample rate (which is generally equal to the
analog-to-digital converter sample rate) and the output sample rate
(which is the channel stimulation rate). A digital signal processor
may implement the peak detector and sample functions according to
the following pseudo-code: TABLE-US-00001 Variables: in: Input
signal (from rectifier). clock: Boolean clock signal indicating
when an output sample should be taken. peak: Internal peak storage
(initialised to 0). out: Output sample. Constants: decay: Fraction
between 0 and 1 that controls the decay time. if (in > peak)
peak = in else peak = peak * decay if (clock) out = peak
[0162] Another prior art method of envelope detection is referred
to as quadrature envelope detection, and a block diagram of this
method is shown in FIG. 8. As is shown, in this scheme, each
frequency channel contains a pair of band-pass filters. The first
filter is known as the in-phase filter (I filter) 60, and may be
the same as that used in the prior art system discussed above. The
second filter is known as the quadrature filter (Q filter) 62, and
has the same magnitude response as the In-phase filter, but its
phase response differs by 90 degrees. If we denote the output of
the In-phase filter as I, and the output of the quadrature filter
as Q, the envelope E is given by: E=square root of
(I.sup.2+Q.sup.2)
[0163] The equivalent signals of this method are shown in FIG. 9
and it is clearly evident that this method produces an envelope
containing no ripple and which is not smeared out, as is the case
in each of the previously described methods. However, this method
does have the disadvantage that it is complex and requires twice
the number of band pass filters than frequency channels.
Furthermore, the implementation of the squaring and square root
operations is not practical in analog circuitry.
[0164] Therefore, the present invention provides a method of
envelope detection that minimises the effect of ripple without
smearing out the envelope and which does not require overly complex
processing that is not practical in analog circuitry, in three
different ways.
[0165] FIG. 10 illustrates the first method of implementing the
system of the present invention, which is applicable for strategies
that use a low sample rate (i.e. channel stimulation rate). The
SPEAK processing strategy developed by the present applicant is
such a strategy, using a channel stimulation rate of 250 Hz. Such
strategies employing a low stimulation rate have the advantage that
power consumption of the system is minimised.
[0166] This method employs an envelope detector comprising a
rectifier 72 and a peak-detect-and-reset stage 74 and is applicable
when the sample rate is low compared to the audio frequencies that
are passed by the band-pass filter 70. Results have shown that in
such instances the present invention acts as an envelope detector
with performance comparable to that of quadrature envelope
detection described above, but with a much lower complexity,
comparable to prior art methods that employ rectifiers and peak
detectors.
[0167] As shown in FIG. 10, the audio signal is passed through a
band pass filter 70, with the resulting signal being passed through
rectifier 72, such as a full-wave rectifier (FWR), as is shown in
FIG. 11. This signal is then processed by a peak-detect-and-reset
stage 74 which combines both the peak detection and sampling
operations. This stage operates in a manner such that when the
received input signal attempts to rise above the output signal, the
output signal follows the input signal, and when the input signal
falls below the output signal, the output signal holds its previous
value (without decay), with the addition that upon the activation
of the clock signal causes the output signal to be sampled and then
momentarily set to zero.
[0168] In the event that the above system uses a FWR, there must be
at least one half-cycle of the FWR output between samples, i.e. the
sample rate must be less than twice the lowest frequency passed by
the BPF. This means that the peak detector always reaches the
maximum value of the FWR output before the next sampling time. For
a constant amplitude audio tone, the output samples will also be
steady in amplitude, without any ripple, aliasing or smearing of
the envelope. The use of a half wave rectifier (HWR) is shown in
FIG. 12, and in this case, it is important that the sample rate is
less than the lowest frequency passed by the BPF.
[0169] One simple embodiment of the present invention is shown in
the circuit 80 of FIG. 13. This embodiment is similar to the prior
art circuit shown in FIG. 6, however the resistor (R) has been
replaced with a reset switch 82. In this embodiment, this reset
switch 82 closes just after the sampling switch 84 opens. In a
two-phase clocking system, the sampling clock (clock 1) to operate
switch 84 would be derived from a phase 1 of the master clock, and
the reset clock (clock 2) to operate switch 82 would be derived
from phase 2 of the master clock.
[0170] It is also possible to implement the peak-detect-and-reset
function of the present invention digitally and this may be done
via a digital signal processor according to the following
pseudo-code: TABLE-US-00002 Variables: in: Input signal (from
rectifier). clock: Boolean clock signal indicating when an output
sample should be taken. peak: Internal peak storage (initialised to
0). out: Output sample. if (in > peak) peak = in if (clock) out
= peak peak = 0
[0171] The second method of implementing the system of the present
invention is to use very high sample rates and as such is
applicable for strategies that use high stimulation rates.
[0172] According to this embodiment of the present invention there
is provided a circuit 90 shown in FIG. 14 wherein each filter 92 is
followed by a half-wave rectifier (HWR) 94, and then sampled by
sample unit 96.
[0173] The phase responses of the individual filters 92 in the
filterbank are designed so that when a pure tone is applied, all of
the filters that pass that frequency have outputs that are in phase
with each other. This condition is readily achieved with
finite-impulse response digital filters. To avoid aliasing, the
sampling rate (i.e. the stimulation rate on that channel) must be
at least four times the highest frequency that is passed by the
band-pass filter. Aliasing is best avoided, because it introduces
spurious frequency components.
[0174] The signals associated with this embodiment of the present
invention are shown in FIG. 15. In this example, the audio input is
a short burst of a 350 Hz pure tone, and the sample rate used is
2000 Hz (stimulation rate of 2000 pulses per second). There are
several options for the electrical stimulation that results when
the HWR output sample is zero. It can result in a stimulus pulse at
the minimum current level (analogous to the CIS strategy).
Alternatively, in a maxima selection strategy, this channel would
not be selected for stimulation in this time interval (analogous to
the ACE strategy).
[0175] To adequately represent the waveform, the sample rate (i.e.
the stimulation rate on that channel) must be much higher than the
highest frequency that is passed by the band-pass filter. Although
the sample rate in this example is more than five times the audio
frequency, it can be seen that the sampling introduces undesirable
ripple. This ripple modulates at a beat frequency, which can give
erroneous pitch cues to the implant recipient. Sample rates of at
least eight times the audio frequency are preferred.
[0176] Another embodiment of this aspect of the present invention
is shown in the circuit 100 of FIG. 16. This embodiment utilises a
peak-detect-and-reset stage 106 instead of a simple sample stage to
reduce the ripple, as described previously. The signals shown
resulting from this embodiment are shown also in FIG. 15.
[0177] As can be seen from the peak detected and reset samples
(bottom illustration in FIG. 15), each group of samples that
represent one of the half cycles of the HWR output now contains at
least one sample that has the same amplitude as the HWR output.
[0178] Yet another alternative embodiment of this aspect of the
present invention is shown in the circuit 110 of FIG. 17. This
embodiment is referred to as half-wave gating (HWG). In FIG. 17, an
envelope signal E is generated by an envelope detector 112
described in any one of the various prior art embodiments
previously. A comparator 114 produces a Boolean gating signal G,
which is high when the BPF 111 output is positive: G=(I>0)
[0179] In other words, if I>0, then G=1 else G=0
[0180] The envelope signal E is then multiplied by the gating
signal through multiplier or mixer 116, that is, the envelope is
on-off modulated by the gating signal: V=G*E
[0181] The signal V is then sampled by sample circuit 118.
[0182] The embodiment of FIG. 17 can be further described with
reference to FIG. 18. In this example, the prior art quadrature
envelope detector as discussed previously in relation to FIG. 8 is
employed. However, that circuit is modified in FIG. 18 by having
the In-phase output from I filter input to the comparator 130 whose
output is fed to multiplier 132. Also, input to multiplier 132 is
the envelop E output from square root function circuit 134. The
output signal from multiplier 132 is then sampled by sampling
circuit 136. The signals associated with this embodiment are shown
in FIG. 19.
[0183] The audio example shown in FIG. 19 is a 350 Hz tone burst as
before. When the audio input has steady amplitude, the non-zero HWG
samples all have the same amplitude, hence there is no amplitude
ripple present. The main advantage of this form is that it does not
need sample rates as high as the earlier HWR scheme, as it is
essentially sampling the envelope signal, which varies more slowly
than the BPF output. A sample rate of four times the highest audio
frequency is adequate in this case, however for the high frequency
channels it may be difficult to achieve a sample rate of at least
four times the highest frequency that is passed by the band-pass
filter. This constraint may be relaxed by using a
peak-detect-and-reset stage 140, as shown in FIG. 20.
[0184] For the specific case of a quadrature envelope detector
being used as the envelope detector, the peak-detect-and-reset
stage 150 shown in FIG. 21 may be used. The signals from this
specific embodiment of FIG. 21 are shown for four different sample
rates in FIG. 22.
[0185] As can be seen, If the sample rate is more than four times
the audio frequency, (in this example 2000 Hz) then the system acts
as a half-wave gating system, and conveys the fine timing content
of the BPF output, giving an enhanced pitch cue to the cochlear
implant recipient. If the sample rate is less than twice the audio
frequency (in this example 500 Hz or 250 Hz), then the system acts
as an envelope detector, as in the prior art. For intermediate
sample rates (in this example 1000 Hz), then it acts as an
imperfect envelope detector, where samples are occasionally
dropped. The perceptual consequences of this are minor, as it is
similar to a timing jitter in the pulses that is sometimes
implemented in the prior art. This system can thus be operated at a
variety of stimulation rates without modification.
[0186] Furthermore, it is known that the phase-locking behaviour of
the auditory nerves is most relevant for low audio frequencies
(below 1000 Hz). Thus it can be advantageous to implement a
cochlear implant system in which a Half-Wave detection scheme is
used for the apical electrode channels (corresponding to low
frequencies), and an envelope detection scheme is used for the
basal electrodes (corresponding to high frequencies). Using a
half-wave gating and peak-detect-and-reset on each channel will
achieve this.
[0187] It may furthermore be advantageous to use a high stimulation
rate on the apical 10 channels (with a half-wave detection scheme)
to obtain enhanced pitch cues; whilst using a lower stimulation
rate on the basal channels (with an envelope detection scheme).
[0188] The third method of implementing the system of the present
invention is a variation of the second method discussed above, but
which addresses a problem associated with pitch perception present
in the second method.
[0189] As discussed above, with reference to FIG. 19, the HWG
samples (bottom waveform) have no amplitude ripple as is desirable
in relation to the problems with prior art methods. The shown HWG
samples essentially consist of bursts of pulses with approximately
50% duty cycle with the rate of burst generally equal to the audio
frequency. However, because the sampling is not synchronised to the
audio frequency, the bursts contain different numbers of pulses, in
the example shown in FIG. 19 most bursts have three pulses but some
bursts have only two pulses. Similarly, the interval between bursts
varies.
[0190] Following a study of five cochlear implant recipients
listening to pure tones processed by the above mentioned system, it
was found that some reported hearing a "warble" in the perceived
sound. In other words, instead of a steady pitch as desired, the
pitch underwent small variations. It is believed that this is due
to the pitch being at least partly dependent upon the interval
between the bursts.
[0191] In this regard, the third method of implementing the present
system is to synchronise the sampling clock with the audio
waveform. FIG. 23 compares the HWR method of FIG. 15 and the HWG
method of FIG. 19 with the third method, however in FIG. 23 the
time scale has been expanded to better show the effects of
synchronisation. The top waveform is the HWR output, before
sampling. The next waveform is a uniform-rate clock. The next
waveform shows the result of sampling the HWR waveform with the
uniform clock which is the same as that shown in relation to FIG.
15. The next waveform shows the result of sampling the HWG waveform
with the uniform clock which is the same as that shown in relation
to FIG. 19.
[0192] The next waveform shows the synchronised clock of the
present method. The synchronised clock consists of bursts of pulses
and within each burst, the pulse rate is the same as the uniform
clock. However, the leading pulse of each burst has been
synchronised to the phase of the band-pass filter output. In this
example, the leading pulse occurs a fixed time interval (one half
of a clock period) after the rising zero crossing of the band-pass
filter output.
[0193] The next waveform (HWR-sync samples) shows the result of
sampling the HWR waveform with the synchronised clock. The HWR-sync
samples occur in bursts of three pulses, and each burst has an
almost identical set of amplitudes, because the samples have
occurred in successive cycles of the band-pass filter output at the
same phases. The HWR-sync samples therefore have much reduced
ripple compared to the earlier HWR samples.
[0194] The final waveform (HWG-sync samples) shows the result of
sampling the HWG waveform with the synchronised clock. The HWG-sync
samples occur in bursts, where each burst has the same number of
pulses (three), and the interval between successive bursts is
constant. Again this is because the samples have occurred in
successive cycles of the band-pass filter output at the same
phases. Five cochlear implant recipients have listened to pure
tones processed by this system, and they all reported a steady
pitch. None of them reported hearing a "warble".
[0195] FIG. 24 shows an implementation of the HWR with clock
synchronisation as described above. Because the synchronised clock
only takes samples when the waveform is positive, the half-wave
rectifier is redundant and can be omitted. FIG. 25 shows an
implementation of the HWG with clock synchronisation described
above.
[0196] Each implementation uses a comparator 160 and clock
synchroniser unit 165, with the latter being enabled only on
positive outputs of the band pass filter using the comparator 160.
In the implementation of FIG. 26 only the in-phase output from the
I filter going positive is used by comparator 160 to enable the
clock synchroniser unit 165.
[0197] The clock synchroniser is shown in FIG. 26. The master clock
is a high frequency clock that determines the overall timing
quantisation. A convenient choice is to set the master clock
frequency equal to the total number of implant stimulation pulses
per second. The enable signal is high when the band-pass filter
output is positive, which then allows the clock to pass through the
AND gate to the divide-by-N stage. This is a simple counter that
outputs one clock pulse each time it receives N clock pulses. The
divide-by-N stage is cleared when Enable is low, so that each new
burst of clock pulses starts on the rising edge of Enable. As an
example, the master clock frequency could be 14400 Hz
(corresponding to a total implant stimulation rate of 14400 pulses
per second), and the divider ratio N could be 8, so that
synchronised clock has a burst rate of 1800 Hz (corresponding to a
channel stimulation rate of 1800 pulses per second).
[0198] This system as described above, assumes that all channels
are independent. This is applicable to a cochlear implant system
that allows simultaneous stimulation on multiple channels. However,
if the cochlear implant system only permits sequential stimulation,
then an additional processing step is required to interleave the
pulses amongst the channels. This is known as an Arbitrator.
[0199] The Arbitrator ensures that on each master clock pulse, no
more than one channel has a low-to-high transition on its Enable
signal. The Enable signals from each channel are the inputs to the
Arbitrator. The Enable signals are passed through the Arbitrator
with controllable delay on each channel. If more than one Enable
signal goes high on a single master clock pulse, then one channel
is selected, and is passed through immediately, but the remaining
channels are delayed by successive master clock periods. In one
embodiment, the Arbitrator prioritises the channels in order from
low frequency to high frequency, so that the low frequency channels
maintain the best synchronisation. In an alternative embodiment,
the channels are prioritised in order of largest to smallest
instantaneous amplitude. More complex rules using both frequency
and amplitude can be devised.
[0200] It is considered that the present invention has significant
advantages over the prior art envelope detection methods discussed
previously. The present invention produces a more accurate envelope
signal from each filter used in the processing strategy without the
complexity and costs associated with quadrature envelope detectors.
The implementation of the present invention will assist cochlear
implant recipients to better perceive rapid changes in the
amplitude of speech, particularly with respect to plosive
consonants. Furthermore, the present invention provides implant
recipients with enhanced pitch cue through the use of a half wave
gating circuit arrangement at high stimulating rates whilst
providing improved envelope detection at low stimulating rates.
[0201] It will be appreciated by persons skilled in the art that
numerous variations and/or modifications may be made to the
invention as shown in the specific embodiments without departing
from the spirit or scope of the invention as broadly described. The
present embodiments are, therefore, to be considered in all
respects as illustrative and not restrictive.
* * * * *