JP3228963B2 - Echo canceller - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a telephone apparatus having, for example, a hands-free communication function, which eliminates an acoustic echo generated by a wraparound of a received voice from a speaker to a microphone and a line echo generated by a hybrid circuit of a communication line. Related to the echo canceller used to

[0002]

2. Description of the Related Art In a communication line, a two-line section such as a subscriber line and a four-line section such as a long-distance trunk line are mixed. Generally, a hybrid transformer is used for a 4-wire to 2-wire converter connecting the 4-wire section and the 2-wire section. However, since the line impedance varies, it is difficult to achieve perfect impedance matching in the hybrid transformer. For this reason, the communication line has four lines.
Line echo occurs in the two-wire converter.

In addition, a telephone device having a so-called hands-free communication function for making a telephone call using a speaker and a microphone provided in a telephone device main body instead of a handset, and a video conference adopting a similar communication form. In the system, the received voice generated from the speaker is reflected on the wall or ceiling and goes around the microphone, so that an acoustic echo is generated.

[0004] These echoes are particularly noticeable in a communication system employing a digital communication system or a communication system having a relatively large transmission delay amount, such as a communication system in which a communication satellite is interposed in a communication line. This is very undesirable because it causes a significant deterioration in quality.

For example, in a digital automobile radio telephone system, a low bit rate speech encoder has begun to be used from the viewpoint of effective use of radio frequency. As a low bit rate speech encoder, for example, CELP (CoP) capable of obtaining relatively good speech quality at 4 to 8 kbps.
de Excited Linear Prediction (VEx), or its improved version VSELP (vector Sun Excited Linear Pr
ediction) method is used. For more details on the CELP method, see MRSchroeder and BSAtal's "Code-E
xcited Linear Prediction (CELP): High-Quality
Speach At Very Low Bit Rates ”in Proc.ICASSP.198
5, pp. 937-939. In these coding systems, generally, coding processing is performed on a frame basis to compress an audio signal to a low bit rate, and interleaving is used to enhance a capability of correcting a burst error. For this reason, the transmission delay in the digital car radio telephone system is about 100 msec one way.

Therefore, conventionally, in this type of system,
A so-called echo canceller is used, which estimates the characteristics of the echo path with an adaptive filter, generates a pseudo echo having the same characteristics as the echo path, and eliminates the echo component contained in the speech signal by subtracting this pseudo echo from the speech signal. Have been. In an automobile radio telephone system, a mobile station is provided with an acoustic echo canceller for canceling an acoustic echo, and a base station is provided with a line echo canceller for canceling a line echo.

FIG. 3 is a circuit block diagram showing an example of the configuration of a digital mobile radio telephone mobile station having an acoustic echo canceller. In the figure, a radio communication signal transmitted from a base station (not shown) using a vacant time slot of a predetermined radio frequency includes an antenna 1 and a duplexer (DU).
The signal is input to a receiving circuit (RX) 3 via a P) 2 and is combined with a local oscillation signal output from a frequency synthesizer (SYN) 4 to be converted into an intermediate frequency signal. The received intermediate frequency signal is supplied to a digital demodulation circuit (D
After the frame synchronization and the bit synchronization by the EM) 6, the digital demodulation is performed. The synchronization signal obtained by the frame synchronization and the bit synchronization is supplied to a control circuit (CONT) 20.

The digital demodulation signal output from the digital demodulation circuit 6 includes a digital communication signal and a digital control signal. Of these, the digital control signal is supplied to the control circuit 20 and identified. On the other hand, the digital speech signal is sampled by the A / D converter 7 and then subjected to error correction decoding by an error correction decoding circuit (CH-DEC) 8. The error-decoded digital call signal is subjected to a decoding process described later in a speech decoding circuit (SP-DEC) 9, and further returned to an analog call signal in a D / A converter 10. The signal is supplied to the speaker 11 and output from the speaker 11.

On the other hand, the transmission signal input by the microphone 12 is sampled by the A / D converter 13 and then input to the voice coding circuit (SP-COD) 15 via the acoustic echo canceller 14 where it is input. Encoded. The digital transmission signal obtained by this encoding is
After being error-correction-coded by an error-correction coding circuit (CH-COD) 16 together with a digital control signal output from the control circuit 20 and further D / A-converted by a D / A converter 17, a digital modulation circuit (MOD) 18 is input. In the digital modulation / demodulation circuit 18, a modulation signal having an intermediate frequency corresponding to the transmission signal supplied from the D / A converter 17 is generated and input to the transmission circuit (TX) 5. In the transmission circuit 5, the modulated signal is combined with a local oscillation signal output from the frequency synthesizer 4 and converted into a high-frequency signal. Then, the high-frequency signal is subjected to high-frequency amplification and transmitted from the antenna 1 to the base station via the duplexer 2.

Reference numeral 21 denotes a console unit (CU) in which a key switch group such as a transmission switch and a dial key and a display are arranged, and 22 denotes a power supply for generating a required operating voltage Vcc based on an output voltage of a battery 23. Circuit.

By the way, the speech decoding circuit 9 is, for example, C
It is composed of an ELP decoder and performs the following decoding processing. That is, the coded speech signal supplied from the error correction decoding circuit 8 is input to the demultiplexer 9a.
The demultiplexer 9a reproduces parameters indicating the characteristics of speech necessary for generating a synthesized speech from the encoded speech signal. The parameters include a linear prediction analysis (LPC:
Linear Predictive Coding) parameter α (i) (i = 1
-10) and pitch period L (i), pitch gain βq (i), codebook number I (i), and codebook gain rq (i) (i = 1 to 5) which are information in subframe units (5 msec).
4).

When each parameter is output from the demultiplexer 9a, the white noise u _{I (i)} (n) corresponding to the code book number I (i) is output from the code book (CB) 9c.
(N = 0 to 39) is read. This white noise u
_{I (i)} (n) is multiplied by a codebook gain rq (i) in a multiplier 9e. In addition, the adaptive codebook (adaptive C
B) 9b outputs a pitch vector b _L (n) (n = 0 to 39) corresponding to the pitch period L (i). This pitch vector b _L (n) is multiplied by a pitch gain βq (i) in a multiplier 9d. The signals output from these multipliers 9e and 9d are mutually added by an adder 9f to become a drive signal r (n) for each subframe. This drive signal r
(n) is expressed as in equation (1). Note that the pitch vector b _L (n) output from the adaptive code book 9b is
It is expressed as in equation (2). Where Lx is the floor function of x that produces the largest integer less than or equal to x.

[0013]

(Equation 1)

[0014]

(Equation 2)

The driving signal r (n) thus generated is L
It is input to the PC synthesis filter 9g. The LPC synthesis filter (LPCFIL) 9g has an LPC parameter α
(i) According to the interpolation LPC parameter α ^* (i) (i = 1 to 10) obtained by linearly interpolating (i = 1 to 10),
It has a transfer function H (Z) represented by the following equation, and according to the transfer function H (Z), a synthesized speech x (n) (n = 0 to 39) corresponding to the drive signal r (n). ) Is output.

[0016]

(Equation 3)

The synthesized speech x (n) output from the LPC synthesis filter 9g is a post-filter (PFIL) 9h
Is input to The post filter 9h is used to improve the perceived quality, and has a transfer function H () represented by the following equation (4) using the interpolated LPC parameter α ^* (i) (i = 1 to 10). Z). The synthesized speech x
(n) is filtered according to the transfer function H (Z), and is output as a synthesized speech y (n) (n = 0 to 39). Note that β and の in equation (4) are 0.5 and 0.8, respectively.
Is used.

[0018]

(Equation 4)

The post filter 9h has the above-mentioned second filter.
In order to correct the slope of the frequency characteristic of the transfer function shown in equation (4), a high-pass filter having the transfer function shown in equation (5) may be connected in cascade. Where u
A value such as 0.5 is used.

[0020]

(Equation 5)

When the received voice decoded by the voice decoding circuit 9 is output from the speaker 11, a part of the received voice is reflected on a wall or a ceiling, and the acoustic echo d
(n) is input to the microphone 12. However, this acoustic echo d (n) is
Is erased as follows. That is, the adaptive filter (adaptive FIL) 14a generates a pseudo echo based on the received voice signal, and the adder 14b subtracts the pseudo echo from the transmitted signal, thereby eliminating the acoustic echo included in the transmitted signal. . At this time, the residual echo that has not been completely eliminated by the adder 14b is input to the adaptive filter 14a. The adaptive filter 14a converts its own transfer function based on the residual echo into a transfer function H of an acoustic echo path.
The tap coefficient is updated so as to approach (Z).

By the way, an FIR type filter is generally used as a conventional adaptive filter because it does not require stability determination and guarantees convergence within certain conditions. The tap coefficient update algorithm includes an algorithm using a least squares method (LS) and an algorithm using a recursive least squares method (RLS). However, from the viewpoint of feasibility, a learning identification method (NLMS) obtained by normalizing the least mean square method (LMS) is often used.
The algorithm based on the learning identification method has an advantage that the amount of operation is relatively small and good characteristics are exhibited. Equation (6) shows that the tap coefficient of the P-order adaptive filter is h
13 shows an update formula of the learning identification method when _j (j = 1 to P).

[0023]

(Equation 6)

However, when the tap coefficients are updated using this kind of algorithm, if the input signal y (n) of the adaptive filter is a signal having no correlation, such as white noise, the tap coefficients are converged. Can be performed at high speed, but when the eigenvalues of the correlation matrix have a strong correlation like speech, the convergence speed of the tap coefficients generally decreases. As a solution to this, an LMS lattice algorithm or the like has been proposed in which an audio signal obtained by linear prediction analysis is whitened and the whitened audio signal is used as an input signal. However, this algorithm requires four times as much computational complexity as the learning identification method, which leaves a problem in its feasibility.

FIG. 4 shows another conventional acoustic echo canceller. 3, the same parts as those in FIG. 3 are denoted by the same reference numerals, and detailed description is omitted. The acoustic echo canceller 14 'employs a series-parallel filter structure, and includes an adaptive filter 14a and an adder 14 of the acoustic echo canceller 14 shown in FIG.
In addition to b, adaptive filters 14c and 14d and an adder 14e are provided.

In the adaptive filter 14c, the A / D converter 1
A pseudo echo is generated based on the transmission voice signal output from No. 3, and the pseudo echo is supplied to the adder 14e. The adder 14e detects a residual echo between the pseudo echo output from the adaptive filter 14c and the pseudo echo output from the adaptive filter 14a.
a and 14c update the tap coefficients to reduce the residual echo. The adaptive filter 14d
Has a transfer function opposite to that of the adaptive filter 14c, and generates a pseudo echo based on the echo output from the adaptive filter 14a according to the tap coefficient supplied from the adaptive filter 14c. . In the adder 14b, the pseudo echo output from the adaptive filter 14d is subtracted from the transmission signal output from the A / D converter 13.

In such a configuration, if the transfer function of the adaptive filter 14a converges to A (Z) and the transfer function of the adaptive filter 14c converges to 1-B (Z), the transfer of the acoustic echo path EC is assumed. For the function H (Z)
Equation (7) holds.

[0028]

(Equation 7)

Therefore, the pseudo filter generated by the application filter 14d having the transfer function 1 / (1-B (Z)) which is the inverse of the transfer function of the adaptive filter 14c and having the same tap coefficients as the adaptive filter 14c. Echo
If the transmission signal H (Z) of the acoustic echo path EC is subtracted by the adder 14b from the transmission signal d (n) output from the A / D converter 13, the IIR echo canceller satisfies the relationship of the expression (8). Is equivalent to being erased. However,
It is necessary to determine the stability of 1 / (1-B (Z)).

[0030]

(Equation 8)

That is, in the serial-parallel type echo canceller, an IIR type echo canceller can be equivalently realized by using an FIR type algorithm. However, also in this case, when the signals input to the adaptive filters 14a and 14c are signals having a strong correlation such as audio signals, there remains a problem that the time required for convergence of the tap coefficients is still long.

[0032]

As described above, an FIR type echo canceller using a simple tap coefficient updating algorithm such as a learning identification method or a series-parallel type echo canceller can be used. However, when the received signal is a signal having a strong correlation such as a voice signal, much time is required for convergence of the tap coefficient, and as a result, there is a problem that a start-up response at the start of a call or the like becomes slow. Was.

The present invention has been made in view of the above circumstances. It is an object of the present invention to have a relatively simple circuit configuration and to enable convergence of tap coefficients in a short time.
Accordingly, it is an object of the present invention to provide an echo canceller which is inexpensive and has excellent response to rising.

[0034]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention relates to an echo provided in a digital voice communication apparatus for decoding and transmitting a speech voice signal by a voice coding / decoding circuit employing a vector quantization system. In the canceller, an inverse filter circuit having a transfer function opposite to the transfer function of the speech codec is provided, and a transmission speech signal including an echo is passed through the inverse filter to remove the correlation by the speech codec. I do. And a transmission voice signal including the echo that has passed through the inverse filter circuit, and a signal including a white noise component generated in a process of decoding the coded voice signal in the voice codec,
For example, a white noise read from a codebook or a signal obtained by multiplying the white noise by a codebook gain,
Tap coefficients are input to the adaptive filter circuits, and the tap coefficients are updated based on the transmitted voice signal and the signal containing the white noise component.

[0035]

As a result, according to the present invention, the tap coefficients of the adaptive filter are updated by the white noise generated by the speech codec and the transmitted speech signal whose correlation has been removed by the speech codec by the inverse filter. It will be performed based on. For this reason, the convergence speed of the tap coefficient is increased,
This makes it possible to improve the voice quality. Also,
Since the white noise generated by the existing speech codec is used, it can be realized with a relatively simple configuration.

[0036]

An embodiment of the present invention will be described below. FIG.
FIG. 2 is a circuit block diagram showing an acoustic echo canceller according to one embodiment of the present invention together with its peripheral circuits. In this figure, the same parts as those in FIGS. 3 and 4 are denoted by the same reference numerals.

First, the audio decoding circuit 90 is configured as follows. That is, the coded speech signal supplied from the error correction decoding circuit is input to the demultiplexer 9a.
The demultiplexer 9a reproduces parameters indicating the characteristics of speech necessary for generating a synthesized speech from the encoded speech signal. The parameters include a linear prediction analysis (LPC:
Linear Predictive Coding) parameter α (i) (i = 1
-10) and pitch period L (i), pitch gain βq (i), codebook number I (i) and codebook gain rq (i) (i = 1 to 5), which are information in subframe units (5 msec).
4).

When each parameter is output from the demultiplexer 9a, the white noise u _{I (i)} (n) corresponding to the code book number I (i) is output from the code book (CB) 9c.
(N = 0 to 39) is read. This white noise u
_{I (i)} (n) is multiplied by a codebook gain rq (i) in a multiplier 9e. In addition, the adaptive codebook (adaptive C
B) 9b outputs a pitch vector b _L (n) (n = 0 to 39) corresponding to the pitch period L (i). This pitch vector b _L (n) is multiplied by a pitch gain βq (i) in a multiplier 9d. The signals output from these multipliers 9e and 9d are mutually added by an adder 9f to become a drive signal r (n) for each subframe. This drive signal r
(n) is expressed as in the above equation (1). The pitch vector b _L (n) output from the adaptive code book 9b
Is expressed as in the above equation (2). Where Lx is the floor function of x that produces the largest integer less than or equal to x.

The driving signal r (n) thus generated is L
It is input to the PC synthesis filter 9g. The LPC synthesis filter (LPCFIL) 9g has an LPC parameter α
(i) The transfer function H () represented by the above equation (3) by the interpolated LPC parameter α ^* (i) (i = 1 to 10) obtained by linearly interpolating (i = 1 to 10). Z), and according to the transfer function H (Z), the drive signal r
A synthesized speech x (n) (n = 0 to 39) corresponding to (n) is output.

The synthesized speech x (n) output from the LPC synthesis filter 9g is a post-filter (PFIL) 9h
Is input to The post-filter 9h is used to enhance the audibility, and has a transfer function P () represented by the following equation (9) by the interpolated LPC parameter α ^* (i) (i = 1 to 10). Z). The synthesized speech x
(n) is filtered according to the transfer function P (Z), and is output as a synthesized speech y (n) (n = 0 to 39). Note that β and の in equation (4) are 0.5 and 0.8, respectively.
Is used.

[0041]

(Equation 9)

The post filter 9h has the above-mentioned second filter.
In order to correct the slope of the transfer function shown in equation (9), a high-pass filter having the transfer function represented by equation (5) may be connected in cascade. Here, a value such as 0.5 is used for u.

On the other hand, the acoustic echo canceller 1 of this embodiment
40 is configured as follows. That is, the acoustic echo canceller 140 of this embodiment differs from the conventional acoustic echo canceller 14 'shown in FIG. 4 in the following point.

(1) The conventional acoustic echo canceller 14 'uses the synthesized speech signal y (n) output from the post filter 9h of the speech decoding circuit 9 as an input signal for the tap coefficient updating algorithm of the adaptive filter. In contrast, in the acoustic echo canceller 140 of this embodiment, the multiplier 9e adds the codebook gain rq (i) to the white noise uI _(i) (n) output from the codebook 9c of the audio decoding circuit 90. ) Is multiplied by a signal rq (i) · u _{I (i)} (n).

(2) The adaptive filter 14
In the conventional acoustic echo canceller 14 ', the acoustic echo signal d (n) output from the A / D converter 13 is used as it is as the echo signal input to c. On the other hand, in the acoustic echo canceller 140 of the present embodiment, an inverse filter group is provided between the A / D converter 13 and the adaptive filter 14c, and the input signal r is provided by the inverse filter group.
An acoustic echo signal d '(n) corresponding to q (i) .u _{I (i)} (n) is created, and this acoustic echo signal d' (n) is used as an input signal of the adaptive filter 14c. .

That is, the acoustic echo signal d (n) output from the A / D converter 13 is
g. The inverse post filter 14g is provided with a transfer function P of the post filter 9h of the audio decoding circuit 90.
It has a transfer function P '(Z) opposite to that of (Z). And
By filtering the acoustic echo signal d (n) using the transfer function P '(Z), the correlation of the acoustic echo signal d (n) by the post filter 9h is removed. Equation (10) shows the transfer function P '(Z).

[0047]

(Equation 10)

When a first-order high-pass filter for correcting the gradient of the transfer function is connected in cascade to the post filter 9h, the inverse filter having the transfer function represented by the equation (11) is inverted. What is necessary is just to cascade connection before the post filter 14g.

[0049]

[Equation 11]

The acoustic echo signal output from the inverse post filter 14g is then converted to an LPC synthesis inverse filter 14h.
Is input to The LPC synthesis inverse filter 14h has a transfer function H ′ (Z) that is opposite to the transfer function of the LPC synthesis filter 9g of the audio decoding circuit 90. And
By filtering the acoustic echo signal according to the transfer function H '(Z), the correlation of the acoustic echo signal by the LPC synthesis filter 9g is removed, and the driving signal r
An acoustic echo signal r '(n) corresponding to (n) is output. Equation (12) shows the inverse transfer function H '(Z).

[0051]

(Equation 12)

The acoustic echo signal r '(n) output from the LPC synthesis inverse filter 14h is equal to the pitch inverse filter 1
4i. The pitch inverse filter 14i generates a pitch vector b corresponding to the acoustic echo signal r '(n).
Output L ′ (n). Equation (13) shows that this pitch vector b
L ′ (n).

[0053]

(Equation 13)

The pitch vector bL '(n) is multiplied by a pitch gain βq (i) in a multiplier 14j, and then supplied as a negative signal to an adder 14k. In the adder 14k, the pitch vector bL '(n) multiplied by the pitch gain βq (i) is subtracted from the acoustic echo signal r' (n) output from the LPC synthesis inverse filter 14h.

[0055] Thus, the input signal _{rq (i) · u I (} i) acoustic echo signal d corresponding to (n) '(n) (the equation (14)) is obtained, input this signal to the adaptive filter 14c Is done. It should be noted that even in the process of processing by the above inverse filter group, L
The PC parameter α ^* (i), the pitch gain βq (i), and the pitch period L (i) change in subframe (5 msec) units as in the speech decoding circuit 90.

[0056]

[Equation 14]

On the other hand, in the adaptive filter group for realizing the tap coefficient updating algorithm, first, the adder 14k
The output signal of the adaptive filter 14c to which the acoustic echo signal d '(n) output from the
Noise rq (i) · u output from the multiplier 9e
An error e '(n) from the output signal of the adaptive filter 14f to which _{I (i)} (n) is input is obtained by the adder 14e. The adaptive filters 14f and 14c use the error e '(n) to update their own tap coefficients h1, j and h2, j to reduce the error e' (n). Here, the tap coefficients h1, j and h2, j are expressed as in equations (15) and (16), respectively.

[0058]

(Equation 15)

[0059]

(Equation 16)

The tap coefficients h1, h1 of the adaptive filter 14f
j is set as a tap coefficient for the adaptive filter 14a. The adaptive filter 14a generates an echo based on the synthesized voice y (n) output from the post filter 9h of the voice decoding circuit 90 according to the tap coefficient, and supplies the echo to the adaptive filter 14d. The adaptive filter 14d has a transfer function (1 / 1-B (Z)) opposite to the transfer function of the adaptive filter 14c, and includes the inverse transfer function and the tap coefficient supplied from the adaptive filter 14c. In response, a pseudo echo is generated based on the echo output from adaptive filter 14a. In the adder 14b, A /
The pseudo echo output from the adaptive filter 14d is subtracted from the transmission signal including the acoustic echo output from the D converter 13.

As described above, in the case of the echo canceller of this embodiment, the white noise rq (i) generated by the speech decoding circuit 90 is used.
A speech decoding circuit 90 using u _{I (i)} (n) and an inverse filter group;
The update operation of the tap coefficients of the adaptive filter group is performed based on the acoustic echo signal d '(n) from which the correlation due to is removed. For this reason, compared with the case where the tap coefficient of the adaptive filter is updated using the output audio signal of the audio decoding circuit and the acoustic echo signal input from the microphone 12, a signal having a small correlation is used. The time required to update the filter tap coefficients is reduced,
Thereby, the convergence speed of the tap coefficient can be increased.

Since the white noise rq (i) · u _{I (i)} (n) generated in the decoding process in the audio decoding circuit 90 is used for updating the tap coefficients, the white noise is newly added. This has the advantage that it can be implemented with a relatively simple circuit configuration and algorithm.

Next, another embodiment of the present invention will be described. FIG. 2 is a circuit block diagram showing the acoustic echo canceller in this embodiment together with its peripheral circuits. The acoustic echo canceller 140 'of the present embodiment is obtained by applying the present invention to the echo canceller shown in FIG. 3 and differs from the circuit of the above embodiment (FIG. 1) only in that the adaptive filter 14c, 14d is omitted.

That is, the white noise rq (i) · u _{I (i)} (n) generated by the speech decoding circuit 90 is input to the adaptive filter 14f. In the adaptive filter 14f, a pseudo echo is generated according to the white noise rq (i) · uI _(i) (n).

On the other hand, the acoustic echo signal d (n) output from the A / D converter 13 first passes through an inverse post filter 14g and an LPC synthesis inverse filter 14h, as in FIG. Is input to The pitch inverse filter 14i outputs the acoustic echo signal r '(n)
Is generated, and a signal obtained by multiplying the pitch vector bL '(n) by the pitch gain βq (i) by the multiplier 14j is used as a negative signal to generate an adder 14k
Is input to In the adder 14k, the above L
The pitch vector bL '(n) multiplied by the pitch gain βq (i) is subtracted from the acoustic echo signal r' (n) output from the PC synthesis inverse filter 14h, and the output signal is input to the adder 14e. Is done. In the adder 14e, the pseudo echo generated by the adaptive filter 14f is subtracted from the acoustic echo signal r '(n) output from the adder 14k, and the residual echo is applied to update the tap coefficients. It is supplied to the filter 14f. The adaptive filter 14f updates the tap coefficient so that the residual echo becomes zero.

The tap coefficients set by the adaptive filter 14f are set as tap coefficients in the adaptive filter 14a. Therefore, the adaptive filter 14a generates a pseudo echo based on the synthesized voice signal output from the voice decoding circuit 90 according to the tap coefficient, and supplies the pseudo echo to the adder 14b. In the adder 14b, the pseudo echo generated by the adaptive filter 14a is subtracted from the transmission voice signal including the acoustic echo output from the A / D converter 13. Thus, a transmission voice signal in which the acoustic echo has been canceled is obtained.

Also in this echo canceller, the tap coefficients in the adaptive filter 14f are changed by the white noise rq (i) · u _{I (i)} (n) generated by the voice decoding circuit 90 and the voice decoding by the inverse filter group. This is performed based on the acoustic echo signal from which the correlation by the circuit has been removed. Therefore, the tap coefficients can be made to converge at a high speed.

The present invention is not limited to the above embodiments. For example, in each of the above embodiments, the multiplier 9e
White noise rq multiplied by the codebook gain rq (i)
(i) · u _{I (i)} (n) is supplied to the adaptive filter 14f, but the white noise u _{I (i)} (n) before being multiplied by the codebook gain rq (i) is applied to the adaptive filter 14f. May be supplied. However, in this case, 1 / rq (i) is added to the acoustic echo signal d '(n) output from the adder 14k.
Need to double.

In each of the above embodiments, the case where the present invention is applied to an acoustic echo canceller has been described. However, the present invention can be similarly applied to a line echo canceller. Further, the present invention can be applied to not only a wireless telephone device but also a telephone device using a wired line.

In addition, the configuration of the adaptive filter circuit and the inverse filter circuit, the type of the tap coefficient updating algorithm, and the like can be variously modified without departing from the gist of the present invention.

[0071]

As described above in detail, according to the present invention, there is provided an inverse filter circuit having a transfer function opposite to the transfer function of the speech codec, and a transmission speech signal including an echo is passed through the inverse filter. This eliminates the correlation by the speech codec. The transmission voice signal including the echo that has passed through the inverse filter circuit and the voice code signal generated by the voice codec circuit provided for decoding the speech voice signal are generated in the process of decoding the coded voice signal. A signal containing a white noise component, for example, white noise read from a codebook or a signal obtained by multiplying this white noise by a codebook gain is input to an adaptive filter circuit, and the transmission voice signal and the white noise The tap coefficient of the adaptive filter circuit is updated based on the signal including the component.

Therefore, according to the present invention, the convergence of tap coefficients can be performed in a short time with a relatively simple circuit configuration, thereby providing an inexpensive echo canceller with excellent start-up response. .

[Brief description of the drawings]

FIG. 1 is a circuit block diagram showing an acoustic echo canceller according to an embodiment of the present invention together with its peripheral circuits.

FIG. 2 is a circuit block diagram showing an acoustic echo canceller according to another embodiment of the present invention together with its peripheral circuits.

FIG. 3 is a circuit block diagram showing an example of a configuration of a digital mobile radio telephone mobile station including a conventional acoustic echo canceller.

FIG. 4 is a circuit block diagram showing a configuration of another conventional acoustic echo canceller.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Antenna, 2 ... Duplexer, 3 ... Receiving circuit, 4 ... Frequency synthesizer, 5 ... Transmission circuit, 6 ... Digital demodulation circuit, 7, 13 ... A / D converter, 8 ... Error correction decoding circuit,
9, 9 ', 90, 90' ... voice decoding circuit, 10, 17, ...
D / A converter, 11 speaker, 12 microphone
14, 14 ', 140, 140': acoustic echo canceller, 15: voice code circuit, 16: error correction code circuit, 1
8 Digital modulation circuit, 20 Control circuit, 21 Console unit, 22 Power supply circuit, 23 Battery, 9a
Demultiplexer, 9b: Adaptive codebook, 9c: Codebook, 9d, 9e, 14j: Multiplier, 9f, 14
b, 14e, 14k: adder, 9g: LPC synthesis filter, 9h: post filter, 14a, 14c, 14d,
14f: adaptive filter, 14g: inverse post filter, 1
4h: LPC synthesis inverse filter; 14i: Pitch inverse filter.

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H04B 3/00-3/44

Claims (3)

(57) [Claims] 1. An echo canceller provided in a digital voice communication apparatus for decoding and transmitting a speech voice signal by a voice codec circuit adopting a vector quantization method, wherein the transfer function is opposite to a transfer function of the voice codec circuit. And an inverse filter circuit for removing the correlation of the transmitted speech signal including the echo by the speech codec, and the transmitted speech signal including the echo that has passed through the inverse filter circuit is input. And decodes speech signals
Noise generated in the process of decoding an encoded audio signal in the audio codec circuit provided for
Is input, and the input transmission voice
An echo canceller comprising: an adaptive filter circuit that updates a tap coefficient based on a signal and a signal including a white noise component . 2. An echo canceller provided in a digital voice communication apparatus for decoding and transmitting a speech voice signal by a voice codec circuit adopting a vector quantization method, wherein the transfer function is opposite to a transfer function of the voice codec circuit. And an inverse filter circuit for removing the correlation of the transmitted speech signal including the echo by the speech codec, and the transmitted speech signal including the echo that has passed through the inverse filter circuit is input. And the speech codec
Input white noise read from the codebook
And an adaptive filter circuit for updating tap coefficients based on the input transmission voice signal and white noise. 3. An echo canceller provided in a digital voice communication device for decoding and transmitting a speech voice signal by a voice coding / decoding circuit employing a vector quantization system, wherein the transfer function is opposite to a transfer function of the voice coding / decoding circuit. And an inverse filter circuit for removing the correlation of the transmitted speech signal including the echo by the speech codec, and the transmitted speech signal including the echo that has passed through the inverse filter circuit is input. And decodes speech signals
The white noise read out from the code book in the audio codec circuit provided for
The signal multiplied by the readbook gain is input and this input
An echo canceller comprising: an adaptive filter circuit that updates tap coefficients based on a transmitted transmission voice signal and a white noise signal .