 |
Claims  |
|
|
We claim:
1. An apparatus for speech processing in a digital telephone system,
comprising:
an echo canceller for receiving a digitized speech-plus-echo signal, for
receiving a far-end speech signal, and for providing an echo-suppressed
output signal, said echo canceller comprising;
state determination means for determining which talk states two speakers
are engaged in, said state determination means providing a state
information signal indicative of said talk states; and
a tone detector responsive to said state information signal.
2. The apparatus of claim 1, wherein said tone detector comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing said digitized speech samples and for providing
a signal indicative of DTMF tone selection and duration;
tone detection means for detecting DTMF tones; and
controller means for disabling said tone detection means when said state
information signal indicates a far-end speech only state.
3. The apparatus of claim 1, wherein said tone detector comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing said digitized speech samples and for providing
a signal indicative of DTMF tone selection and duration;
tone detection means for detecting DTMF tones; and
controller means for enabling said tone detection means when said state
information signal indicates a near-end speech only state.
4. In a speech processing apparatus comprising an echo canceller and a
digital processing element, a method for controlling the operation of said
digital processing element using state information from said echo
canceller, the method comprising the steps of:
generating a state information signal by said echo canceller indicative of
a plurality of talk states; and
controlling a tone detector function within said digital processing element
using said state information signal.
5. The method of claim 4 wherein the controlling step is further comprised
the steps of:
disabling said tone detector function within said digital processing
element when said state information signal indicates far-end speech only.
6. The method of claim 4 wherein the controlling step is further comprised
the steps of:
enabling said tone detector function within said digital processing element
when said state information signal indicates near-end speech only.
7. An apparatus for speech processing in a digital telephone system,
comprising:
an echo canceller for receiving a digitized speech-plus-echo signal, for
receiving a far-end speech signal, and for providing an echo-suppressed
output signal, said echo canceller comprising;
state determination means for determining which talk states two speakers
are engaged in, said state determination means providing a state
information signal indicative of said talk states; and
a noise suppressor, separate from said echo canceller, responsive to said
state information signal.
8. The apparatus of claim 7, wherein said noise suppressor comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing a noise-suppressed digitized speech signal;
background noise estimation means for generating an estimated background
noise signal used to suppress background noise; and
controller means for disabling said background noise estimation means when
said state information signal indicates far-end only speech.
9. The apparatus of claim 7, wherein said noise suppressor comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing a noise-suppressed digitized speech signal;
background noise estimation means for generating an estimated background
noise signal used to suppress background noise; and
controller means for enabling said background noise estimation means when
said state information signal indicates both speakers silent.
10. The apparatus of claim 7, wherein said speech processing means is a
noise suppressor, comprising of:
input means for receiving digitized speech samples, for receiving said
state information signal, and for receiving said signal indicative of
whether or not echo is present at said input to said echo canceller;
output means for providing a noise-suppressed digitized speech signal;
background noise estimation means for generating an estimated background
noise signal used to suppress background noise; and
controller means for enabling said background noise estimation means when
said state information signal indicates far-end only speech and said echo
detection means indicates no echo present.
11. An apparatus for speech processing in a digital telephone system,
comprising:
an echo canceller for receiving a digitized speech-plus-echo signal, for
receiving a far-end speech signal, and for providing an echo-suppressed
output signal, said echo canceller comprising;
state determination means for determining which talk states two speakers
are engaged in, said state determination means providing a state
information signal indicative of said talk states; and
a transmission muting means, separate from said echo canceller, responsive
to said state information signal.
12. The apparatus of claim 11 wherein said transmission muting means
comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing either said digitized speech samples or a
synthesized noise signal;
noise generation means for generating said synthesized noise signal; and
controller means for replacing said digitized speech samples with said
synthesized noise when said state information signal indicates far-end
only speech.
13. An apparatus for speech processing in a digital telephone system,
comprising:
an echo canceller for receiving a digitized speech-plus-echo signal, for
receiving a far-end speech signal, and for providing an echo-suppressed
output signal, said echo canceller comprising;
state determination means for determining which talk states two speakers
are engaged in, said state determination means providing a state
information signal indicative of said talk states; and
a vocoder encoder, separate from said echo canceller, responsive to said
state information signal.
14. The apparatus of claim 13, wherein said vocoder encoder comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing an encoded digital speech packet at a reduced
data rate;
background noise estimation means for generating threshold information used
to determine which rate to encode said digitized speech samples; and
controller means for disabling said background noise estimation means when
said state information signal indicates far-end speech only.
15. The apparatus of claim 13, wherein said vocoder encoder comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing an encoded digital speech packet at a reduced
data rate;
background noise estimation means for generating threshold information used
to determine which rate to encode said digitized speech samples; and
controller means for enabling said background noise estimation means when
said state information signal indicates both speakers silent.
16. An apparatus for speech processing in a digital telephone system,
comprising:
an echo canceller for receiving a digitized speech-plus-echo signal, for
receiving a far-end speech signal, and for providing an echo-suppressed
output signal, said echo canceller comprising;
state determination means for determining which talk states two speakers
are engaged in, said state determination means providing a state
information signal indicative of said talk states; and
an adaptive equalizer, separate from said echo canceller, responsive to
said state information signal.
17. The apparatus of claim 16, wherein said adaptive equalizer comprises:
input means for receiving digitized speech samples and for receiving said
state information signal;
output means for providing a frequency compensated digitized speech signal;
frequency estimation means for estimating the spectral content of said
digitized speech samples; and
controller means for enabling said frequency estimation means when said
state information signal indicates far-end only speech.
18. In a speech processing apparatus comprising an echo canceller and a
digital processing element, a method for controlling the operation of said
digital processing element using state information from said echo
canceller, the method comprising the steps of:
generating a state information signal by said echo canceller indicative of
a plurality of talk states; and
controlling a noise suppressor function, separate from said echo canceller,
within said digital processing element using said state information
signal.
19. The method of claim 18 wherein the controlling step is further
comprised the steps of:
disabling a background noise estimation calculation performed by said noise
suppression function within said digital processing element when said
state information signal indicates far-end speech only.
20. The method of claim 18 wherein the controlling step is further
comprised the steps of:
enabling a background noise estimation calculation performed by noise
suppression function within said digital processing element when said
state information signal indicates both speakers silent.
21. The method of claim 18, wherein the controlling step is further
comprised of:
disabling a background noise estimation calculation performed by said noise
suppression function within said digital processing element when both said
state information signal indicates far-end speech only and said echo
present signal indicates echo present at input to said echo canceller; and
enabling said background noise estimation calculation when both said state
information signal indicates far-end speech only and said echo present
signal indicates no echo present at input to said echo canceller.
22. In a speech processing apparatus comprising an echo canceller and a
digital processing element, a method for controlling the operation of said
digital processing element using state information from said echo
canceller, the method comprising the steps of:
generating a state information signal by said echo canceller indicative of
a plurality of talk states; and
controlling a transmission muting function, separate from said echo
canceller, within said digital processing element using said state
information signal.
23. The method of claim 22 wherein the controlling step is further
comprised the steps of:
enabling said transmission muting function which replaces digitized speech
with synthesized noise when said state information signal indicates
far-end speech only and disabling said transmission muting function for
all other talk states.
24. In a speech processing apparatus comprising an echo canceller and a
digital processing element, a method for controlling the operation of said
digital processing element using state information from said echo
canceller, the method comprising the steps of:
generating a state information signal by said echo canceller indicative of
a plurality of talk states; and
controlling a vocoder encoder function, separate from said echo canceller,
within said digital processing element using said state information
signal.
25. The method of claim 24 wherein the controlling step is further
comprised the steps of:
disabling a background noise estimation calculation performed by said
vocoder encoder function within said digital processing element when said
state information indicates far-end speech only.
26. The method of claim 24 wherein the controlling step is further
comprised the steps of:
enabling a background noise estimation calculation performed by said
vocoder encoder function within said digital processing element when said
state information signal indicates both speakers silent.
27. In a speech processing apparatus comprising an echo canceller and a
digital processing element, a method for controlling the operation of said
digital processing element using state information from said echo
canceller, the method comprising the steps of:
generating a state information signal by said echo canceller indicative of
a plurality of talk states; and
controlling an adaptive equalizer function, separate from said echo
canceller, within said digital processing element using said state
information signal.
28. The method of claim 27 wherein the controlling step is further
comprised the steps of:
enabling a frequency response update performed by said adaptive equalizer
function within said digital processing element when said state
information signal indicates near-end speech only. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to digital telephone systems. More
particularly, the present invention relates to a novel and improved method
and apparatus for using state determination from an echo canceller to
control various functional blocks in a digital telephone system.
2. Description of the Related Art
Transmission of voice by digital techniques has become widespread,
particularly in cellular telephone and PCS applications. This, in turn,
has created an interest in improving speech processing techniques. Three
of such techniques include the addition of echo cancellers, noise
suppressors, and voice encoders/decoders, or vocoders, to existing
elements of digital telephone systems.
Echo cancellers are used to diminish undesired echo signals caused by
impedance mismatches in land-based telephone networks, or in the case of
mobile telephones, echo caused by acoustic coupling between speaker and
microphone in "hands free" telephones. Vocoders are used to remove natural
redundancies of speech in a digitized signal in order to reduce data
transmission rates and consequently the amount of information being
transmitted over a given transmission channel. Noise suppressors are used
to minimize background noise. Echo cancellers, vocoders, and noise
suppressors are presently used together in digital telephone systems both
in land-based applications and in mobile systems.
There are two types of echo cancellers, the network echo canceller and the
acoustic echo canceller. An example of a typical network echo canceller is
disclosed in U.S. Pat. No. 5,307,405 entitled "NETWORK ECHO CANCELLER",
which is assigned to the assignee of the present invention and
incorporated by reference herein. A network echo canceller cancels the
echo produced in a telephone network. A land-based telephone is connected
to a central office by a two wire line to support transmission in both
directions. For calls farther than about 35 miles, the two directions of
transmission must be segregated onto physically separate wires, resulting
in a four-line wire. The device that interfaces the two-wire and four-wire
segments is known as a hybrid. An impedance mismatch at the hybrid results
in an echo which must be removed by a network echo canceller. Acoustic
echo cancellers are used in teleconferencing and hands-free telephony
applications. An acoustic echo canceller eliminates acoustic echo
resulting from the feedback between a loudspeaker and a microphone.
In a typical digital telephone system, speech is converted from an analog
signal to digital PCM samples by an A/D converter. In a typical
embodiment, a data rate of 64 kbps is chosen in order to retain good voice
quality. Once the speech signal has been digitized, it can be manipulated
to achieve certain benefits, such as maximization of system capacity,
speech quality enhancement, noise suppression, and minimization of
transmission errors.
After the speech signal has been converted to PCM samples, undesired echo
can be removed by an echo canceller, background noise can be minimized by
a noise suppressor, and data compression can be performed by a vocoder
before modulation and upconversion for transmission. An example of a
variable rate vocoder is disclosed in U.S. Pat. No. 5,414,796 entitled
"VARIABLE RATE VOCODER", which is assigned to the assignee of the present
invention and incorporated by reference herein. The encoded speech signal
can be modulated by any number of techniques, including TDMA, CDMA, or
analog modulation. The use of CDMA techniques in a multiple access
communication system is disclosed in U.S. Pat. No. 4,901,307, entitled
"SPREAD SPECTRUM MULTIPLE ACCESS COMMUNICATION SYSTEM USING SATELLITE OR
TERRESTRIAL REPEATERS," which is assigned to the assignee of the present
invention and incorporated by reference herein. Combining the echo
canceller with the vocoder and noise suppressor has certain benefits as
well as problems associated with it.
One problem with introducing an echo canceller into the front end
electronics of a digital telephone system is that it alters the speech
signal to the other functional blocks due to its location in the system
relative to the other functional blocks. By placing the echo canceller
first in the chain of functional blocks, the noise suppressor and vocoder
must make background noise calculations based on an echo-canceled signal
rather than actual background noise. If the echo canceller does not remove
all of the echo from the speech signal, the residual echo can cause errors
in the background noise calculations performed by the noise suppressor and
vocoder.
Herein, a mobile user is referred to as the near-end speaker and the
land-based user is referred to as the far-end speaker. A typical vocoder
may contain a noise suppressor whose function is to remove background
noise from the near-end speech signal. An example of a typical noise
suppressor is disclosed in U.S. Pat. No. 4,811,404 entitled "NOISE
SUPPRESSION SYSTEM", which is assigned to Motorola, Inc. and incorporated
by reference herein. Noise suppression is performed by calculating an
estimate of the actual background noise energy during periods when the
near-end speaker is silent. A problem occurs if the near-end speaker is
silent and the far-end speaker is talking. In the mobile telephone, the
far-end speaker's voice can be acoustically coupled from the speaker to
the microphone, resulting in an echo that will be heard by the far-end
speaker unless it is removed. In a land-based system, near-end speech can
be coupled onto the far-end speaker's voice signal due to the impedance
mismatch in the hybrid discussed above. An echo canceller is used to
eliminate the echo, but because of limitations of the echo canceller, the
echo will not be completely removed. A noise suppressor placed after the
echo canceller may interpret the residual echo as background noise and
update the background noise estimate based upon the residual echo. This
corrupts the background noise estimate, resulting in degraded noise
suppression. The vocoder will suffer by providing a poor estimate of the
background noise to a synthesized noise generator in the system. In
addition, the vocoder's encoding rate decisions will be adversely
affected.
It is therefore an object of the present invention to prevent erroneous
background noise updates in the noise suppressor and the vocoder encoder
when the near-end speaker is silent and the far-end speaker is active.
It is another object of the present invention to use the state
determination signal from the echo canceller to control other functional
elements within a digital telephone system, such as a tone detector, a
transmission mute function, and an adaptive equalizer.
SUMMARY OF THE INVENTION
The present invention is a novel and improved combination of functional
elements within a digital telephone system. In accordance with the present
invention, an echo canceller is used in combination with a vocoder wherein
the echo canceller provides information to various functional blocks
within the vocoder for purposes of noise suppression, DTMF tone detection,
transmission muting and voice encoding. An immediate benefit of combining
the an echo canceller with a vocoder is the cost, weight, and space
savings of combining two integrated circuits onto a single integrated
circuit.
In the exemplary embodiment of the present invention, an echo canceller is
used that determines, among other things, which mode of speech that two
persons are engaged in. In the exemplary embodiment, five different modes
or talk states are possible: near-end speech only, far-end speech only,
both speakers talking, neither speaker talking, and hangover, which is the
brief period of time immediately following a pause in the conversation.
The present invention uses the state determination of the echo canceller in
several functional blocks within the vocoder. Of particular importance is
the use of the state determination signal in the noise suppressor function
within the vocoder. In the exemplary embodiment, the noise suppressor
operates by dividing the input signal into selected frequency bands,
generating a signal-to-noise ratio for each frequency band, then
amplifying each frequency band according to a pre-defined gain table. The
speech/noise determination is performed as follows. The raw
signal-to-noise ratio estimates for each frequency band are used to index
a voice metric table to obtain voice metric values for each channel. A
voice metric is a measurement of the overall voice-like characteristics of
the channel energy. The individual channel voice metric values are summed
to create a multi-channel energy parameter, and then compared to a
background noise update threshold. If the voice metric sum does not meet
the threshold, the input frame is deemed to be noise, and a background
noise update is performed. If the voice metric sum exceeds the threshold,
then that frame is treated as speech and the background noise estimate is
not updated. Problems can occur if the noise suppressor treats residual
echo from the echo canceller as background noise. In that case, the noise
estimation algorithm will re-calculate the background noise based on the
residual echo, which would corrupt the noise estimate.
The present invention eliminates this problem by providing a state
information signal from the echo canceller which disables background noise
updates in the noise suppressor when the talk mode is determined to be
far-end only. Without the state information from the echo canceller, the
noise suppressor will erroneously update the background noise calculation
based on the residual echo signal from the echo canceller.
In an alternative embodiment, a second signal from the echo canceller is
provided to the noise suppressor indicating whether any echo is in fact
present at the input to the echo canceller. The second signal will allow
background noise estimates to be performed in the noise suppressor if no
echo is present at the echo canceller, even when the state information
signal would otherwise disable the update.
Furthermore, in the present invention, the state determination from the
echo canceller is used to control the tone detector function within the
vocoder. The tone detector checks the transmission signal for DTMF tones.
If tones are detected, the normal transmission signal is muted and a
signaling message is sent over-the-air that causes the tones to be
generated at the receiver. This is done because a sufficiently high
erasure rate can degrade a vocoded tone enough that it will not be
detected. The tone detector can be disabled by the state determination
signal from the echo canceller during the far-end only talk state,
resulting in power savings.
In addition, the present invention uses state determination from the echo
canceller to control the transmission mute function within the vocoder.
The transmission mute replaces PCM samples with synthesized noise that
matches the spectral characteristics of the actual background noise. The
spectral information and volume control for the synthesized noise is
provided by analysis performed by the vocoder encoder. The transmission
mute function is enabled when the echo canceller's state determination
indicates far-end speech only. This way, all echo is eliminated from the
transmission signal.
The present invention also uses state determination from the echo canceller
to control an adaptive equalizer. This equalizer modifies the frequency
response of the received near-end signal to compensate for frequency
response degradations in the transmission path. The equalizer estimates
the frequency characteristics of the transmission path during near-end
speech and uses this estimation to construct a filter that shapes the
overall frequency response to a desired characteristic. Since this
estimate of the received frequency response would be corrupted by the
presence of an echo signal, the echo canceller only allows the equalizer
to update its estimate of the frequency response during the near-end only
speech state.
Finally, the present invention uses state determination from the echo
canceller to control the background noise estimate function that is
performed by the vocoder encoder. This background noise estimate is
performed in order to generate synthesized noise information to be used by
the transmission mute block, discussed above, and to generate threshold
information used to decide which data rate to encode with. The goal is to
match the synthesized noise with the actual noise-suppressed background
noise so that the far-end listener is oblivious to periods of synthesized
noise replacement. The background noise calculation is enhanced by
providing state information from the echo canceller to the background
noise estimate function. The echo canceller disables the background noise
estimate during periods of synthesized noise replacement so that a
background noise update is not performed on synthesized noise.
BRIEF DESCRIPTION OF THE DRAWINGS
The features, objects, and advantages of the present invention will become
more apparent from the detailed description set forth below when taken in
conjunction with the drawings in which like reference characters identify
correspondingly throughout and wherein:
FIG. 1 is a functional block diagram of a mobile digital telephone;
FIG. 2 is a functional block diagram of an echo canceller and a vocoder;
FIG. 3 is a functional block diagram of an echo canceller;
FIG. 4 is a functional block diagram of a noise suppressor;
FIG. 5 is a functional block diagram of a tone detector;
FIG. 6 is a functional block diagram of a transmission mute processor; and
FIG. 7 is a functional block diagram of a vocoder encoder.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
FIG. 1 is an overall block diagram of a digital cellular or PCS telephone.
For simplicity of explanation, only a subset of elements are shown. The
digital cellular telephone consists of handset 6, which includes
microphone 4 and speaker 2; Analog-to-Digital (A/D) converter 8; echo
canceller 10; vocoder 12; transceiver 14; and antenna 16. It should be
understood that other architectures may be employed for the system with a
mere change in location or position of the various operational elements.
During transmission, near-end speech is received by microphone 4 provided
in handset 6. The near-end speech signal is transformed by microphone 4
into an electro-acoustic signal represented by the term v(t) as shown in
FIG. 1. Received far-end speech signal x(t) is acoustically coupled to
speech signal v(t) at summer 5, modeled as passing x(t) through unknown
echo channel 7 to produce echo signal y(t). The output of summer 5 is
shown as combined speech/echo signal v(t)+y(t). Unknown echo channel 7 and
summer 5 are not included elements in the system itself, but rather are
parasitic results from the physical proximity of microphone 4 and speaker
2.
The speech/echo signal v(t)+y(t) is then converted from an analog signal to
PCM samples by Analog-to-Digital converter 8. In an exemplary embodiment,
PCM samples are output by A/D converter 8 at a rate of 64 kbits per second
and are represented by signal s(n) as shown in FIG. 1.
Echo canceller 10 removes echo signal y(t) from digitized speech/echo
signal s(n). In the exemplary embodiment, echo canceller 10 operates in
accordance with the echo canceller described in aforementioned U.S. Pat.
No. 5,307,405. In the exemplary embodiment, echo canceller 10 performs
echo cancellation by determining which of several different talk states
the speakers are engaged in, the states being near-end speech only,
far-end speech only, both near and far-end speech simultaneously, neither
speaker talking, or hangover. Once the talk state is determined by echo
canceller 10, an estimate of echo signal y(n) is removed from digitized
speech/echo signal s(n). Because the echo signal cannot be completely
eliminated, a residual echo signal will remain as part of the digitized
speech signal. This echo canceled speech signal, s'(n), is then processed
by vocoder 12. In the exemplary embodiment, vocoder 12 is a variable rate
code excited linear prediction (CELP) vocoder as described in
aforementioned U.S. Pat. No. 5,414,796. In the exemplary embodiment,
vocoder 12 operates in conjunction with a noise suppression system as
described in detail in aforementioned U.S. Pat. No. 4,811,404.
Vocoder 12 performs several functions on signal s'(n) including, but not
limited to, speech compression, noise suppression, transmit and receive
volume control, DTMF tone detection, and transmission muting. In the
present invention, vocoder 12 uses the state determining results from echo
canceller 10, shown as "state information" in FIG. 1, in its algorithm to
decide when to update its background noise estimate. Further details of
echo canceller 10 and vocoder 12 are shown in FIG. 2 and are discussed
more fully later herein.
The vocoded speech signal, s"(n), is then provided to transceiver 14 where
it is modulated in accordance with a predetermined modulation format such
as Code Division Multiple Access (CDMA), Time Division Multiple Access
(TDMA), Frequency Division Multiple Access (FDMA), or analog modulation.
In the exemplary embodiment, transceiver 14 modulates the signal in
accordance with a CDMA modulation format as described in the
aforementioned U.S. Pat. No. 4,901,307. Transceiver 14 then upconverts and
amplifies the modulated signal. The modulated signal is then transmitted
through antenna 16 to base station transceivers (not shown).
A similar reciprocal process occurs for received speech. A CDMA modulated
signal is received at antenna 16 and provided to transceiver 14.
Transceiver 14 amplifies, downconverts, and demodulates the received
signal. In the exemplary embodiment, transceiver 14 demodulates the
received signal in accordance with a CDMA demodulation format as described
in the aforementioned U.S. Pat. Nos. 5,103,459 and 4,901,307. The
demodulated signal, z"(n), is provided to vocoder 12.
In the exemplary embodiment, vocoder 12 receives encoded variable-length
data packets every 20 ms at data rates ranging from 1200-9600 bps. Vocoder
12 decodes the packets into 64 kbps PCM samples according to the
aforementioned U.S. Pat. No. 5,414,796. Then the decoded signal, z'(n), is
provided to echo canceller 10 where it is used as a reference to remove
undesired echo signal y(t) from the desired speech signal. The decoded
signal output from echo canceller 10 is shown as z(n) in FIG. 1.
Finally decoded signal z(n) is converted to an analog waveform by A/D
converter 8, then converted to acoustic far-end speech using speaker 2
provided in handset 6.
FIG. 2 is a functional block diagram of echo canceller 10 and vocoder 12.
In an exemplary embodiment, echo canceller 10 and vocoder 12 are
configured in digital processor form, such as model ADSP-2181 of the
ADSP-2100 series of digital signal processors manufactured by Analog
Devices of Norwood, Mass. It should be understood that other digital
signal processors may be programmed to function in accordance with the
teachings herein. Alternatively, other implementations of echo canceller
10 and vocoder 12 may be configured from discrete processors or in
application specific integrated circuit (ASIC) form. It should also be
understood that vocoder 12 can be configured using any combination of
functional blocks shown in FIG. 2.
During transmission, digitized speech/echo signal s(n) is received by Tx
PCM Filters 52 from A/D converter 8. Low frequency components are filtered
out because echo canceller 10 cannot synthesize a DC component. The
filtered signal is provided to summer 32 within echo canceller 10 where
estimated echo signal y'(n) is subtracted from it. The estimated echo
estimate signal, y'(n), is produced by processing received digital speech
signal z'(n) using an adaptive filtering operation performed within echo
canceller 10. An example of echo canceller 10 is disclosed aforementioned
U.S. Pat. No. 5,307,405. The details of echo canceller 10 will be
described in greater detail later herein.
The output produced by echo canceller 10 contains the desired digitized
speech signal plus a residual signal left over from the echo canceling
process. The residual signal will be present because the echo canceller
can never completely remove all of the echo from the digitized speech
signal.
The output signal is then provided to tone detector 34, where it is checked
to see if the signal contains DTMF tones. If the signal contains DTMF
tones, Tx Mute 42 is activated by tone detector 34 and transceiver 14 is
instructed to send DTMF tone signals. To save computation, tone detector
34 is bypassed if echo canceller 10 determines that the talk state is
far-end only or both speakers silent.
In the phone, the output signal from echo canceller 10 is then processed by
noise suppressor 38 which attenuates heavy background noise.
Alternatively, in the base station, an adaptive equalizer is used in place
of noise suppressor 38 to dynamically control the frequency content of the
digitized speech signal from the near-end user. An example of an adaptive
equalizer is disclosed in co-pending U.S. patent application Ser. No.
08/953,102, which is a file wrapper continuation of Ser. No. 08/456,277
filed Apr. 28, 1995, entitled "METHOD AND APPARATUS FOR PERFORMING
ADAPTIVE EQUALIZATION", assigned to the assignee of the present invention
and incorporated by reference herein. An example of noise suppressor 38 is
disclosed in aforementioned U.S. Pat. No. 4,811,404. It should be
understood that other implementations of noise suppressor 38 may be used
other than the one disclosed in U.S. Pat. No. 4,811,404.
Noise suppressor 38 updates its estimate of the background noise
characteristics by measuring the spectral characteristics of the incoming
signal. The present invention provides a state determination signal from
echo canceller 10 to aid in the background noise estimate update decision.
Allowing the echo canceller to assist in the enabling and disabling of the
background noise estimate update provides significant advantages that will
become obvious later herein.
The noise-suppressed speech signal from noise suppressor 38 is then
provided to Tx mute 42 which, when enabled, replaces the digitized speech
signal with synthesized noise which in the exemplary embodiment matches
the spectral characteristics of the actual background noise. If Tx Mute 42
is disabled, the speech signal is provided to vocoder encoder 44
unchanged. Tx Mute 42 is enabled by echo canceller 10 during the far-end
only talk state.
The speech signal is then passed from Tx mute 42 to vocoder encoder 44. An
example of vocoder encoder 44 and vocoder decoder 46 is disclosed in
aforementioned U.S. Pat. No. 5,414,796. In the exemplary embodiment,
vocoder encoder 44 accepts digitized speech samples at 64 kbps and
compresses it to achieve a reduced data rate. This is accomplished by
removing all of the natural redundancies inherent in speech. The basis of
this technique is to compute the parameters of a filter, called the LPC
filter, which performs short-term predictions of the speech waveform using
a model of the human vocal tract. In addition, long term effects, related
to the pitch of the speech, are modeled by computing the parameters of a
pitch filter, which essentially model the human vocal chords. Finally,
these filters must be excited, and this is done by determining which one
of a number of random excitation waveforms in a codebook results in the
closest approximation to the original speech when the waveform excites the
two filters mentioned above. A background noise estimation is also
performed within vocoder encoder 44 which estimates the energy of the
background noise during periods of silence. Since the background noise
estimate should only be updated on actual background noise, it is
desirable to use the state information signal from echo canceller 10 to
determine when both near-end and far-end speakers are silent. Without this
information from echo canceller 10, the background noise estimate may be
updated even when synthesized noise is supplied by Tx mute 42, which is
undesirable. Further details of vocoder encoder 44 will be provided later
herein.
In the receiving direction, again referring to FIG. 2, data is accepted
from transceiver 14 and processed by vocoder decoder 46. In the exemplary
embodiment, vocoder decoder 46 accepts variable-length data packets at
data rates ranging from 1200 to 9600 bps or from 1200 to 13000 bps and
produces 64 kbps PCM samples according to aforementioned U.S. Pat. No.
5,414,796 and is shown as z'(n). These PCM samples are then routed through
echo canceller 10 to A/D converter 8. Z'(n) is also used by echo canceller
10 as a reference signal to cancel the echo in the Tx direction. The
output of echo canceller 10 in the Rx direction is shown as z(n).
To better understand the present invention, a working knowledge of the
various functional blocks is needed. FIG. 3 is a detailed block diagram of
echo canceller 10. An example of echo canceller 10 is disclosed in
aforementioned U.S. Pat. No. 5,307,405. It should be understood that in
the exemplary embodiment, echo canceller 10 is in essence a state machine
that has defined functions for each of the five different talk states
described above.
In FIG. 3, as was for FIG. 2, the speech signal from the mobile station is
labeled as the near-end speech s(n), while the far-end speech signal from
Rx PCM filters 50 is labeled as z'(n). Z'(n) is amplified by variable gain
stage 170 and coupled to s(n) at summer 5, modeled as passing through
unknown echo channel 7. To remove low-frequency background noise, the sum
of the echo signal y(n) and the near-end speech signal s(n) is high-pass
filtered by Tx PCM filters 52 to produce signal r(n). Signal r(n) is
provided as one input to each of summers 32 and 150, and control unit 152.
The input far-end speech z'(n) is fed to variable gain stage 170 and then
stored in buffer 154 for input to a set of transversal adaptive filters
(initial filter 156, state filter 158 and echo canceller filter 160), and
control unit 152.
During the period of normal operation of echo canceller 10, signal y.sub.1
(n) is output from state filter 158 to one input of summer 150 where it is
subtracted from signal r(n). The resultant output from summer 150 is the
signal e.sub.1 (n) which is input to control unit 152. The output of echo
canceller filter 160, the echo replica signal y(n) , is provided through
filter switch 162 to one input of summer 32 where it is subtracted from
the signal r(n). The resultant echo residual signal e(n) output from
summer 32 is fed back as an input to control unit 152. Echo residual
signal e(n) as output from summer 32 may be provided directly as the
output of the echo canceller 10, shown as s'(n), or through additional
processing elements, not shown.
To prevent large background noise levels from interfering with state
determination, echo canceller 10 performs a differential energy algorithm
on signals z'(n) and e(n). This algorithm continually monitors the
background noise level and compares it with the signal energy to determine
if the speaker is talking. Three thresholds, T.sub.1 (B.sub.i), T.sub.2
(B.sub.i), and T.sub.3 (B.sub.i), are first calculated which are functions
of the background noise level B.sub.i. If the signal energy of the signal
x(n) exceeds all three thresholds, the speaker is determined to be
talking. If the signal energy exceeds T1 and T2 but not T3, the speaker is
determined to be probably uttering an unvoiced sound, such as the "sp"
sound in the word "speed." If the signal energy is smaller than all three
thresholds, the speaker is determined to be not talking.
As illustrated in FIG. 3, two independently-adapting filters, filters 158
and 160, track the unknown echo channel. While filter 160 performs the
actual echo cancellation, filter 158 is used by the control unit 152 to
determine which of several states echo canceller 10 should be operating
in. This state information is provided to various functional blocks within
vocoder 12, including tone detector 34, noise suppressor/adaptive
equalizer 38, Tx mute 42, and vocoder encoder 44.
FIG. 4 is a functional block diagram of noise suppressor 38. An example of
noise suppressor 38 is disclosed in aforementioned U.S. Pat. No.
4,811,404. It should be understood that other implementations of noise
suppressor 38 may be used other than the one disclosed in U.S. Pat. No.
4,811,404. The noise suppression system includes a mechanism 210 for
separating the input signal into a plurality of pre-processed signals
representative of selected frequency channels; a mechanism 310 for
generating an estimate of the signal-to-noise ratio (SNR) in each
individual channel; a mechanism 830 for calculating the noise energy in
each frequency channel; a mechanism 590 for producing a gain value for
each individual channel by automatically selecting one of a plurality of
gain values from a particular gain table in response to the channel SNR
estimates; a mechanism 250 for modifying the gain of each of the plurality
of pre-processed signals in response to the selected gain values to
provide a plurality of post-processed noise-suppressed output signals; and
a mechanism 260 for combining the post-processed signals back into time
domain PCM data. Voice metric calculator 810 is used to perform the
speech/noise decision making process. First, the raw SNR estimates from
channel SNR estimator 310 are used to index a voice metric table to obtain
voice metric values for each channel. A voice metric is a measurement of
the overall voice-like characteristics of the channel energy. The
individual channel voice metric values are summed to create a first
multi-channel energy parameter, and then compared to a background noise
update threshold in threshold comparator 820. If the voice metric sum does
not exceed the threshold, the input frame is deemed to be noise, and a
background noise update is performed by enabling noise energy calculator
830 to re-calculate the noise energy in each channel. The estimated noise
energy is used by gain table 590 to select the appropriate gain for each
channel. If the voice metric sum exceeds the update threshold, that frame
is deemed to be a voice frame, and noise energy calculator 830 is disabled
from updating the noise energy estimate. The present invention provides an
additional enable signal from echo canceller 10 that disables noise energy
calculator 830 when echo canceller 10 determines that only far-end speech
is occurring. This enable signal takes precedence over the enable signal
from threshold comparator 820; that is, if noise energy calculator 830 is
disabled by the signal from echo canceller 10, it will remain disabled
even when provided | | |
