JP4652116B2 - Echo canceller - Google Patents

Description

  The present invention relates to an echo canceller that reduces acoustic echoes output from a speaker and input to a microphone via an echo path.

When it is assumed that there is no input sound other than echo when inputting the sound reproduced by the stereo speaker to the microphone, the transmission input signal s _in (n) input to the microphone can be expressed by the following equation (1). . However, the received signal time series vector of the Left channel is indicated by bold l _in (n) (for the purposes of electronic application, the emphasized alphabet letter is expressed as bold alphabet), and the received signal time series vector of the Right channel is indicated by bold r _in (n) and then, Left side and Right side speakers - each bold h _L the impulse response of the transfer function between microphone and bold h _R. Note that N is N> 0 and is sufficiently large with respect to the echo reverberation time. Further, the bold h _L bold l _in (n) and the bold h _R bold r _in (n) represent the echo by the Left channel and the echo by the Right channel, respectively.

The conventional stereo echo canceller adaptively learns the bold h _L and bold h _R using an adaptive filter with l _in (n) as input and an adaptive filter with r _in (n) as input. Estimated value of impulse response of transfer function between side speaker and microphone A bold h _L hat and a bold h _R hat are obtained. Then, bold h _L hat bold l _in (n) and bold h _R hat bold r _in (n) are generated as pseudo echo signals from each estimated value, and these pseudo echo signals are generated from the transmission input signal s _in (n). Echo is eliminated by pulling off.

  As a technique for performing such stereo echo cancellation, for example, there is a technique disclosed in Patent Document 1.

JP 2001-1000078 A

  The conventional stereo echo canceller has a problem that the echo canceling effect is lowered when the received signal is frequently switched between stereo and monaural.

FIG. 9 is a diagram for explaining a stereo sound recording method, and the above problem will be described in more detail with reference to FIG. In the example of FIG. 9, a stereo sound signal is recorded using two microphones for one sound source. The sound o (n) generated from the sound source inputs the influence of the propagation characteristics of the space on the path propagated until reaching the Left and Right microphones. In this case, the sound waveform of the sound collected by the Left and Right microphones can be expressed by the following equation (2). However, the bold p _L (n) and the bold p _R (n) are transfer functions of the space on the path propagated until reaching the Left and Right microphones.

In reality, it is rare that bold p _L (n) = bold p _R (n), and l _in (n) ≠ r _in (n) holds _in normal stereo sound. Thus, the audio signal normally recorded in stereo can be regarded as a separate signal in which the L channel and the R channel do not completely coincide with each other even if they are correlated with each other. For this reason, in the conventional stereo echo canceller, it is possible to estimate the bold letter h _L and the bold letter h _R which are impulse responses of the transfer function.

However, if bold l _in (n) = bold r _in (n), it is impossible to distinguish between the Left channel echo and the Right channel echo included _in s _in (n). Thus, it is impossible to estimate the bold characters h _L and h _R.

In such a case, s _in (n) has the relationship of the following formula (3).

Thus, both adaptive filters that receive l _in (n) or r _in (n) learn and identify (bold h _L + bold h _R ). Therefore, in this case, the conventional stereo echo canceller learns and identifies (bold h _L + bold h _R ) simultaneously with two adaptive filters.

As described above, in a stereo audio signal, it can be said that bold l _in (n) = bold r _in (n) is rare, whereas a monaural audio signal is always bold l _in (n) = bold r _in ( n). In the case of receiving both stereo sound and monaural sound, the conventional stereo echo canceller changes the learning and identification target of the adaptive filter when switching from stereo to monaural or from monaural to stereo.

For example, when the received signal changes from stereo speech to monaural speech, the target of learning identification is changed from bold h _L and bold h _R to (bold h _L + bold h _R ). The values of bold h _L hat and bold h _R hat, which are values, are forgotten and new (bold h _L + bold h _R ) must be learned and identified again.

  At this time, during the period from the start of relearning to the time of convergence, the effect of removing the echo is lower than usual, and the residual echo is increased. If the received signal is frequently switched between stereo and monaural, it is expected that the echo canceling effect by the stereo stereo echo canceling device is reduced accordingly.

  Further, when the conventional stereo echo canceller is used for monaural sound, there are the following problems.

As described above, the conventional stereo echo canceller has one adaptive filter that receives l _in (n) and r _in (n) as input, but monaural speech has bold l _in (n) = bold r _{Since in} (n), the same signal is input to each adaptive filter. Assuming that the coefficient sequence of each adaptive filter is bold a ₁ (n), bold a ₂ (n) and the output of the adaptive filter is d ₁ (n), d ₂ (n), these are given by the following equation (4): It is expressed in Therefore, the residual signal s _out (n) obtained by subtracting the output d ₁ (n), d ₂ (n) of the adaptive filter from the transmission input signal s _in (n) can be expressed by the following equation (5).

Here, when bold l _in (n) = bold r _in (n) = bold u (n), s _in (n) is expressed by the following equation (6), and s _out (n) is expressed by the following equation: It becomes like (7). However, the bold _letter h _sum and the bold _letter a _sum (n) are in the relationship of the following formula (8).

Also, the coefficient update formula at this time is expressed by the following formula (9) when the NLMS algorithm is used, for example. Where α is a coefficient update step gain. Furthermore, in the following formula (9), when the bold letters a ₁ (n) and bold letters a ₂ (n) are put together, the following formula (10) is obtained.

  Thus, from the above equations (7) and (10), two parallel adaptive filters having the same input signal behave the same as one adaptive filter having a double coefficient update step gain. I understand that.

  In general, when the coefficient update step gain of an adaptive filter is increased in an echo canceller, there is an advantage that followability is improved by a change in the propagation path of the echo, but there is a disadvantage that the amount of echo cancellation is not stable even after the coefficient has converged. . This is because as the amount of correction of the coefficient increases, it becomes easier to input the influence of noise and speaker speech components included in the transmission signal, and the adaptation error increases.

  On the other hand, when the coefficient update step gain is reduced, the echo cancellation amount after the coefficient convergence is stabilized, but the followability to changes in the echo path is reduced. Therefore, the conventional stereo echo canceller has a problem that the adaptation error becomes larger than that in the case of stereo sound because the effective coefficient update step gain increases when monaural sound is input.

  On the other hand, if the coefficient update step gain is set to be small in order to suppress the adaptation error at the time of monaural sound input, the follow-up to the change of the echo path at the time of stereo sound input is delayed.

  As described above, the conventional stereo echo canceller cannot obtain a sufficient echo cancel effect for a signal in which the coincidence of the Left channel signal and the Right channel signal of the received sound changes every moment.

  Such a tendency is expected to be a problem when, for example, an echo due to stereo sound such as music or TV sound is erased by a stereo echo erasing device. In addition, as an application scene of the echo canceller, for example, in a TV or music player equipped with a speech recognition device as in Patent Document 1, an echo such as a TV sound or a reproduced music sound superimposed on the recognized speech is eliminated by stereo echo. The case where it erases with an apparatus is mentioned.

  In the case of music and TV audio, even in the case of stereo audio content, monaural audio may be superimposed or inserted in the middle, and the state in which stereo audio is dominant and the state in which monaural audio is dominant are from moment to moment. And change. In particular, TV sound changes in stereo sound and monaural sound even when the viewing channel is changed or the program changes.

  As described above, in the conventional stereo echo device, the echo canceling effect changes depending on the match between the left channel signal and the right channel signal of the received voice of the received voice, so that the performance is not always stable in the above-described application example. Cannot be obtained.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide an echo canceling apparatus that can perform stable echo canceling regardless of the monaural nature of a received signal.

The echo canceling apparatus according to the present invention is adapted to receive and input 2- channel audio signals, analyze the degree of coincidence between the received signals of each channel, and extract a pseudo-echo signal from the received signals of 2 channels. A first adaptive filter set composed of the adaptive filter, a second adaptive filter set composed of an adaptive filter corresponding to the extraction of the pseudo echo signal of the monaural signal obtained from the two- channel received signal , A threshold is set for the degree of coincidence between received signals. When the degree of coincidence is lower than the threshold, the first adaptive filter set is selected, and when the degree of coincidence is higher than the threshold, the second adaptive filter set is selected. From the transmission input signal using the filter set selection means and the adaptive filter set selected by the filter set selection means Those comprising echo canceling means for outputting as a transmission output signal except echo signals similar.

  According to the present invention, a reception signal analyzing means for receiving and inputting a multi-channel audio signal and analyzing the degree of coincidence between the reception signals of each channel, and an adaptive filter corresponding to the extraction of a pseudo echo signal of the multi-channel reception signal The first adaptive filter set composed of the second adaptive filter set composed of the adaptive filter corresponding to the extraction of the pseudo echo signal of the monaural signal obtained from the multi-channel received signal, and the degree of coincidence between the channels. A filter set selection unit that selects one of the adaptive filter sets according to the analysis result and an adaptive filter set selected by the filter set selection unit removes the pseudo echo signal from the transmission input signal and outputs the transmission output signal. Echo canceling means, so even if you receive a multi-channel audio signal Regardless of the mono property, there is an effect that it is possible to perform a stable echo cancellation.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing the configuration of an echo canceling apparatus according to Embodiment 1 of the present invention. The echo cancellation apparatus according to the first embodiment includes a first adaptive filter set 101, a second adaptive filter set 102, a received signal analysis unit (received signal analysis unit) 103, and a filter set selection unit (filter set selection unit) 104. A filter set selection means comprising: a first coefficient update control section (first coefficient update control means) 105; a second coefficient update control section (second coefficient update control means) 106; a subtractor ( (Subtracting means) 107, a subtractor 108, and an echo canceling means having an output signal switching device (output signal switching means) 109.

The first adaptive filter set 101 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n) and applies the sum of the pseudo echo signals obtained by applying these signals to the adaptive filter. Is output as a pseudo echo signal d ₁ (n). The second adaptive filter set 102 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n) and applies the sum signal as a monaural signal to the adaptive filter to generate a second pseudo echo signal. d ₂ (n) is output.

The received signal analysis unit 103 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and outputs a received signal analysis result J (n) relating to the degree of coincidence of both signals. The filter set selection unit 104 outputs a filter set selection result SEL (n) for selecting one of the adaptive filter sets 101 and 102 according to the input received signal analysis result J (n).

The first coefficient update control unit 105 receives the filter set selection result SEL (n) and controls the coefficient update of the first adaptive filter set 101 according to SEL (n). The second coefficient update control unit 106 receives the filter set selection result SEL (n) and controls the coefficient update of the second adaptive filter set 102 according to SEL (n). The subtractor 107 receives the transmission input signal s _in (n) and the first pseudo echo signal d ₁ (n), and outputs a first residual signal s ₁ (n) that is the difference between them.

The subtractor 108 inputs the transmission input signal s _in (n) and the second pseudo echo signal d ₂ (n), and outputs a second residual signal s ₂ (n) that is the difference between them. The output signal switch 109 receives the first residual signal s ₁ (n), the second residual signal s ₂ (n), and the filter set selection result SEL (n) output from the filter set selection unit 104. The residual signal corresponding to the filter set selected in SEL (n) is output as a transmission output signal s _out (n).

Next, the operation will be described.
First, the received signal analysis unit 103 analyzes the degree of coincidence between the left channel received signal time series vector bold l _in (n) and the right channel received signal time series vector bold r _in (n). The analysis result is output as a received signal analysis result J (n). Here, the degree of coincidence between the bold letter l _in (n) and the bold letter r _in (n) is evaluated by an evaluation function such as the following formula (11), for example. In the following equation (11), the power ratio of the sum signal and the difference signal of the bold letter l _in (n) and the bold letter r _in (n) is represented. In the following formula (11), the value of J (n) increases as the matching degree between bold l _in (n) and bold r _in (n) increases, and conversely, the value of J (n) decreases as the matching degree decreases. Becomes smaller. In addition, you may evaluate using an evaluation function like following formula (12).

The above equation (12) represents the correlation between bold l _in (n) and bold r _in (n). The higher the matching degree, the larger the value of J (n), and the lower the matching degree. The value of J (n) becomes smaller.

The filter set selection unit 104 selects an adaptive filter set to be used by threshold determination using a predetermined threshold for the reception signal analysis result J (n) output from the reception signal analysis unit 103. When J (n) falls below a predetermined threshold, the degree of coincidence between the bold l _in (n) that is the Left channel received signal time series vector and the bold r _in (n) that is the Right channel received signal time series vector is low. It can be judged. Therefore, the first adaptive filter set corresponding to this is selected as the adaptive filter set to be used, and SEL (n) = 1 is output.

Further, when J (n) exceeds a predetermined threshold value, it can be determined that the matching degree between bold l _in (n) and bold r _in (n) is high, so the second adaptive filter set corresponding to this is selected. SEL (n) = 2 is output. If it is on the boundary between the two, either adaptive filter set may be selected. In addition, if information useful for judging the degree of coincidence between bold l _in (n) and bold r _in (n) can be obtained from the outside separately for the received signal, SEL ( n) may be determined.

  The filter set selection unit 104 outputs the filter set selection result SEL (n) determined as described above to the first coefficient update control unit 105, the second coefficient update control unit 106, and the output signal switch 109. .

  The first coefficient update control unit 105 is connected to the first adaptive filter set 101, and in accordance with the adaptive filter set selection result SEL (n) output from the filter set selection unit 104, the first coefficient update control unit 105 Controls execution or stop of coefficient update processing. That is, when SEL (n) = 1, the coefficient update process of the first adaptive filter set 101 is executed, and when SEL (n) = 2, the coefficient update process of the first adaptive filter set 101 is stopped.

  Similarly to the first coefficient update control unit 105, the second coefficient update control unit 106 also uses the second adaptive filter set 102 according to the adaptive filter set selection result SEL (n) output from the filter set selection unit 104. The execution or stop of the coefficient update process is controlled. That is, when SEL (n) = 2, the coefficient update process of the second adaptive filter set 102 is executed, and when SEL (n) = 1, the coefficient update process of the second adaptive filter set 102 is stopped.

  It should be noted that other conditions such as those used in a general echo canceller can be added to the decision to execute or stop the coefficient update of the adaptive filter. The present invention does not exclude these possibilities. For example, a double talk detector may be provided, and the double talk detection result may be added to the condition for executing or stopping the coefficient update.

The first adaptive filter set 101 and the second adaptive filter set 102 are adaptive filter sets composed of one or more adaptive filters. The first adaptive filter set 101 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and outputs the first pseudo echo signal d ₁ (n) based on these input signals. To do. The second adaptive filter set 102 receives l _in (n) and r _in (n) and outputs a second pseudo echo signal d ₂ (n) based on these input signals.

The configuration of the first adaptive filter set 101 and the second adaptive filter set 102 may be, for example, as shown in FIG. In FIG. 2, the first adaptive filter set 101 includes a first adaptive filter 201, a second adaptive filter 202, and an adder (adding unit) 203. The first adaptive filter 201 receives the Left channel reception signal l _in (n), and receives the Left channel reception by the bold a ₁ (n) that is the first adaptive filter coefficient held in the first adaptive filter 201. The signal time series vector bold l _in (n) is filtered, and the Left channel component pseudo echo signal d _L1 (n) is output.

Also, the second adaptive filter 202 receives the right channel received signal r _in (n), and the right adaptive filter 202 holds the right by the bold a ₂ (n) that is the second adaptive filter coefficient. Channel received signal time series vector bold r _in (n) is filtered, and a Right channel component pseudo echo signal d _R1 (n) is output.

That is, the first adaptive filter 201 and the second adaptive filter 202 perform filtering according to the following equation (13), and output pseudo echo signals d _L1 (n) and d _R1 (n), respectively.

The adder 203 receives the Left channel component pseudo echo signal d _L1 (n) and the Right channel component pseudo echo signal d _R1 (n), adds them according to the following equation (14), and adds the first pseudo echo signal d. ₁ Output (n).
d ₁ (n) = d _L1 (n) + d _R1 (n) (14)

The second adaptive filter set 102 includes a sum signal generator 204 and a third adaptive filter 205, and the sum signal generator 204 includes the Left channel received signal l _in (n) and the Right channel received signal. r _in (n) is input, and the received sum signal q _S (n) is output according to the following equation (15). The received sum signal q _S (n) is a monaural signal obtained from the Left channel received signal l _in (n) and the Right channel received signal r _in (n).
q _S (n) = {l _in (n) + r _in (n)} / 2 (15)

The third adaptive filter 205 receives the received sum signal q _S (n), and filters the bold q _S (n) according to the following equation (16) by the bold a ₃ (n) that is the third adaptive filter coefficient. The second pseudo echo signal d ₂ (n) is output. Subtractor 107 receives the transmission input signal s _in (n) and the first pseudo echo signal d ₁ (n), subtracts the d ₁ (n) from s _in (n) according to the following equation (17), Output as the first residual signal s ₁ (n). The subtractor 108 receives the transmission input signal s _in (n) and the second pseudo echo signal d ₂ (n), and subtracts d ₂ (n) from s _in (n) according to the following equation (18). And output as the second residual signal s ₂ (n).

Here, the first adaptive filter 201 and the second adaptive filter 202 included in the first adaptive filter set 101 use the first residual signal s ₁ (n) and the first adaptive filter coefficient. A certain bold character a ₁ (n) and a second adaptive filter coefficient bold character a ₂ (n) are updated. Further, in the third adaptive filter 205 included in the second adaptive filter set 102, the second adaptive filter coefficient Bold a ₃ (n) is updated using the second residual signal s ₂ (n). To do.

A general algorithm such as the NLMS algorithm can be used as the adaptation algorithm used for coefficient update. As an example, the following formula (19) is shown as a coefficient update formula when the NLMS algorithm is used. Here, α ₁ , α ₂ , and α ₃ are the adaptive filter coefficient update step gains.

  Whether or not coefficient update is executed depends on the controls 105 and 106 of the coefficient update control unit connected to each adaptive filter set.

The first residual signal s ₁ (n) and the second residual signal s ₂ (n) are input to the output signal switch 109. The output signal switch 109 outputs a residual signal corresponding to the selected filter set as a transmission output signal s _out (n) according to the filter set selection result SEL (n) output from the filter set selection unit 104. That is, when SEL (n) = 1, s ₁ (n) is output as s _out (n), and when the second adaptive filter set 102 is selected, s ₂ (n) is transmitted. Output as signal s _out (n).

  As described above, according to the first embodiment, when the degree of coincidence between the Left channel signal and the Right channel signal of the received signal, that is, when the monaural property is low, echo cancellation is performed by the first adaptive filter set 101, which is high. In this case, echo cancellation is performed by the second adaptive filter set 102. Here, the configuration of the first adaptive filter set 101 may be the same as that of the conventional stereo echo canceller, and the configuration of the second adaptive filter set 102 may be the same as that of the conventional monaural echo canceller.

  Thereby, even when the monaurality (stereoness) of the received signal changes from moment to moment, an appropriate adaptive filter configuration can be selected according to the monaurality (stereoness) of the received signal. Therefore, it is possible to avoid the problem of increase in residual echo when switching between stereo and monaural, and the problem of adaptation error when monaural sound is input, as seen in conventional stereo echo cancellers. Echo cancellation effect is obtained.

  In addition, an effect of stabilizing the echo cancellation effect can be obtained regardless of the degree of coincidence between the Left channel signal and the Right channel signal of the received signal.

Further, according to the first embodiment, it is possible to suppress the influence of the fluctuation of the matching degree between the Left channel signal and the Right channel signal of the received signal on the echo cancellation effect.
In addition, the filter set selection unit shown in the first embodiment and the later-described embodiment selects a filter set according to the degree of coincidence between the reception signals of each channel described above, and based on an external control signal You may comprise so that a filter set can be selected.

Embodiment 2. FIG.
A conventional stereo echo canceller apparatus cancels echoes using an adaptive filter that inputs l _in (n) and an adaptive filter that inputs r _in (n). On the other hand, in the second embodiment, echo is canceled using an adaptive filter that inputs a sum signal of l _in (n) and r _in (n) and an adaptive filter that receives a difference signal. is there.

When the above formula (1) is rewritten, the following formula (20) can be derived.

From this equation, it is clear that s _in (n) is the sum of the linear responses of the bold letters h _S and the bold letters h _D and the bold letters q _S (n) and q _D (n) as transfer functions. Accordingly, the bold q _s (n) is similar to the case where it is possible to learn and identify the bold characters h _L and h _R which are transfer functions by l _in (n) and r _in (n) and erase the echo. It is also possible to eliminate the echo by learning and identifying the bold characters h _S and h _D of the transfer function using the bold characters q _D (n).

  FIG. 3 is a block diagram showing a configuration of an echo cancellation apparatus according to Embodiment 2 of the present invention. The echo cancellation apparatus according to the third embodiment includes a sum signal generator 301, a difference signal generator 302, a first adaptive filter 303, a second adaptive filter 304, an adder 305, and a subtractor 306.

The sum signal generator 301 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and obtains and outputs the received sum signal q _S (n) from these signals. The difference signal generator 302 receives l _in (n) and r _in (n), and obtains and outputs a received difference signal q _D (n) from these signals. The received sum signal q _S (n) and the received difference signal q _D (n) are monaural signals obtained from the received signals l _in (n) and r _in (n). The first adaptive filter 303 receives the received sum signal q _S (n) and outputs a first pseudo echo signal d ₁ (n). The second adaptive filter 304 receives the reception difference signal q _D (n) and outputs a second pseudo echo signal d ₂ (n).

Next, the operation will be described.
The sum signal generator 301 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and outputs the received sum signal q _S (n) according to the above equation (15).

The first adaptive filter 303 receives the received sum signal q _S (n), and filters the bold q _S (n) according to the following formula (21) with the bold a ₁ (n) that is the first adaptive filter coefficient. The obtained signal is output as a sum signal component pseudo echo d _S (n). The difference signal generator 302 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and outputs a received difference signal q _D (n) according to the following equation (22).

The second adaptive filter 304 receives the reception difference signal q _D (n), and filters q _D (n) according to the following equation (23) with the boldface a ₂ (n) that is the first adaptive filter coefficient. The signal is output as a difference signal component pseudo echo d _D (n). The adder 305 inputs the sum signal component pseudo echo d _S (n) and the difference signal component pseudo echo d _D (n), and adds d _S (n) and d _D (n) according to the following equation (24). To output a pseudo echo signal d (n). The subtractor 306 receives the transmission input signal s _in (n) and the pseudo echo signal d (n), subtracts d (n) from s _in (n) according to the following equation (25), and transmits the transmission output signal s _out. (N) is output.

In addition, the first adaptive filter 303 and the second adaptive filter 304 use s _out (n) to update the coefficients of the bold letters a ₁ (n) and bold letters a ₂ (n), which are the respective adaptive filter coefficients. When the NLMS algorithm is used, the coefficient update formula is as shown in the following formula (26).

As described above, the stereo echo canceller according to the second embodiment includes the adaptive filter that receives the sum signal q _S (n) of l _in (n) and r _in (n), and the difference signal q _D (n ) Is used to cancel the echo. When the received signal is a monaural audio signal, that is, in the relationship of the following formula (27), q _S (n) can be expressed by the following formula (28), and q _D (n) is expressed by the following formula (29). There is a relationship.

As a result, since the input signal q _D (n) to the second adaptive filter is a silent signal, the coefficient update of the adaptive filter is performed only for the boldface a ₁ (n) that is the first adaptive filter coefficient. Done.

  In the conventional stereo echo canceller, when the received signal is a monaural audio signal, there is a problem that the echo canceling performance is deteriorated by inputting the same signal to the two adaptive filters. On the other hand, in the echo canceller of the second embodiment, even if the received signal is a monaural audio signal, the input signals of the two adaptive filters are different, and thus the above-described problem is avoided. As a result, even when a monaural audio reception signal is input, an effect that the echo cancellation capability is not deteriorated can be obtained.

  Further, according to the second embodiment, since the configuration is simple compared to the configuration in which the adaptive filter set is replaced according to the received signal, an effect that the apparatus can be reduced in size can be obtained.

Embodiment 3 FIG.
In the echo canceller according to Embodiment 1 described above, echo cancellation is performed using one adaptive filter that receives the sum signal component of a received signal having a high degree of coincidence between the Left channel received signal and the Right channel received signal. In the third embodiment, echo cancellation is performed by further adding an adaptive filter that receives the difference signal component.

  FIG. 4 is a diagram showing a configuration example of an adaptive filter set in the echo cancellation apparatus according to Embodiment 3 of the present invention. The echo cancellation apparatus according to the third embodiment includes a first adaptive filter set 41 including a first adaptive filter 401, a second adaptive filter 402, and an adder 403, a sum signal generator 404, and a difference signal generation. 405, a third adaptive filter 406, a fourth adaptive filter 407, and a second adaptive filter set 42 comprising an adder 408. Other components are the same as those of the echo canceling apparatus of the first embodiment, and the description thereof is omitted.

In the first adaptive filter set 41, the first adaptive filter 401 receives the Left channel received signal l _in (n), and the bold l _in (n) is the bold a ₁ (n) that is the first adaptive filter coefficient. ) To output a Left channel component pseudo echo signal d _L (n). The second adaptive filter 402 receives the right channel received signal r _in (n), filters the bold r _in (n) with the bold a ₂ (n) that is the second adaptive filter coefficient, and outputs the right channel. The component pseudo echo signal d _R (n) is output. The adder 403 receives d _L (n) and d _R (n), adds them according to the following equation (30), and outputs a _first pseudo echo signal d ₁ (n).

In the second adaptive filter set 42, the sum signal generator 403 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and receives the received sum signal q _S according to the above equation (15). (N) is output. The reception sum signal q _S (n) is a monaural signal obtained from the reception signals l _in (n) and r _in (n).

The third adaptive filter 406 receives the received sum signal q _S (n), and filters the bold q _S (n) according to the following formula (31) by the bold a ₃ (n) that is the third adaptive filter coefficient. The sum signal component pseudo echo signal d _S (n) is output. Here, the bold q _s (n) is a signal time series vector including past data.

The difference signal generator 405 receives the Left channel received signal l _in (n) and the Right channel received signal r _in (n), and outputs a received difference signal q _D (n) according to the following equation (32).

The fourth adaptive filter 407 receives the received difference signal q _D (n), and filters the bold q _D (n) according to the following equation (33) by the bold a ₄ (n) that is the fourth adaptive filter coefficient. The difference signal component pseudo echo signal d _D (n) is output. Note that bold q _D (n) is a signal time-series vector including past data.

The adder 408 inputs d _S (n) and d _D (n) and outputs a second pseudo echo signal d ₂ (n). The first adaptive filter 401, a second adaptive filter 402, a third adaptive filter 406, an adaptive filter coefficient of the fourth adaptive filter 407, bold a ₁ (n), bold a ₂ (n), bold a _If the coefficient update expression of ₃ (n) and bold a ₄ (n) is an NLMS algorithm, for example, the following expression (34) is obtained.

  As described above, according to the third embodiment, for a received signal having a high match between the Left channel signal and the Right channel signal, an adaptive filter that inputs the sum signal component and a difference signal component are input. Therefore, even when the degree of coincidence between the Left channel signal and the Right channel signal of the received signal is high, not only the echo due to the sum signal component of the Left channel signal and the Right channel signal but also the echo due to the difference signal component The effect of erasing is obtained.

Embodiment 4 FIG.
In the echo canceling apparatus according to the first embodiment, two adaptive filter sets are provided, and these are operated in parallel, and an output signal is switched and used separately. One adaptive filter set is provided, and the adaptive filters are substantially exchanged by appropriately exchanging the adaptive filter coefficients.

  FIG. 5 is a block diagram showing a configuration of an echo cancellation apparatus according to Embodiment 4 of the present invention, and FIG. 6 is a diagram showing a detailed configuration example of echo cancellation means in the echo cancellation apparatus according to Embodiment 4. . The echo cancellation apparatus according to the fourth embodiment includes a filter set selection unit including a received signal analysis unit 501 and a filter set selection unit 502, a filter set control unit 503, a reference input signal generation unit 504, and an adaptive filter set 505. And echo canceling means having a subtractor 506.

The received signal analysis unit 501 receives the received signals l _in (n) and r _in (n) and outputs the received signal analysis result J (n). The filter set selection unit 502 inputs the received signal analysis result J (n) and outputs the filter set selection result SEL (n).

The filter set control unit 503 receives the filter set selection result SEL (n), and controls the update process of the adaptation coefficient for the adaptive filter in the adaptive filter set 505. The reference input signal generation unit 504 receives the filter set selection result SEL (n) and outputs the first reference input signal q ₁ (n) and the second reference input signal q ₂ (n). The adaptive filter set 505 receives reference input signals q ₁ (n) and q ₂ (n) and outputs a pseudo echo signal d (n). The subtractor 506 receives the transmission dominant signal s _in (n) and the pseudo echo signal d (n) and outputs a transmission output signal s _out (n).

  The operations of reception signal analysis section 501 and filter set selection section 502 are the same as those of reception signal analysis section 103 and filter set selection section 104 in the first embodiment, and description thereof is omitted.

The filter set control unit 503 receives the filter selection result SEL (n), and controls the adaptive filter set 505 at the transition of SEL (n). The reference input generation unit 504 outputs reference input signals q ₁ (n) and q ₂ (n) suitable for the selected filter set according to the filter set selection result SEL (n). These functions will be described later.

The adaptive filter set 505 receives q ₁ (n) and q ₂ (n) and outputs a pseudo echo signal d (n). The subtractor 506 receives the transmission input signal s _in (n) and the pseudo echo signal d (n), and subtracts d (n) from s _in (n) according to the above equation (25) to eliminate the echo. Output signal s _out (n) is output.

Next, details of the echo canceling means according to the fourth embodiment shown in FIG. 6 will be described.
In the example of FIG. 6, the reference input signal generation unit 504 includes, for example, a sum signal generator 601, a difference signal generator 602, a signal switch 603, and a signal switch 604. The sum signal generator 601 receives the received signals l _in (n) and r _in (n) and outputs the sum signal q _S (n) according to the above equation (15).

The signal switch 603 receives l _in (n) and q _S (n) and outputs a reference input signal q ₁ (n) according to the filter selection result SEL (n). That is, when SEL (n) = 1, since the selected adaptive filter set is an adaptive filter set having l _in (n) and r _in (n) as reference input signals, l _in (n) Is output as the first reference input signal q ₁ (n). When SEL (n) = 2, since it is an adaptive filter set having q _S (n) and q _D (n) as reference input signals, q _S (n) is output as q ₁ (n).

On the other hand, the difference signal generator 602 receives the received signals l _in (n) and r _in (n) and outputs a difference signal q _D (n) according to the following equation (35).

The signal switch 604 inputs r _in (n) and q _D (n), and in accordance with the filter selection result SEL (n), when SEL (n) = 1, r _in (n) is input to the second reference. The signal q ₂ (n) is output. When SEL (n) = 2, q _D (n) is output as q ₂ (n).

In the example of FIG. 6, the adaptive filter set 505 includes a first adaptive filter 605, a second adaptive filter 606, and an adder 607. The first adaptive filter 605 receives the first reference input signal q ₁ (n) and stores q ₁ (n) as buffer data bold b ₁ (n) in the signal buffer in the first adaptive filter 605. Then, the buffer data bold b ₁ (n) is filtered according to the following equation (36) with the first adaptive filter coefficient bold a ₁ (n), and the first pseudo echo signal d ₁ (n) is filtered. Output. Note that b _{1, j} (n) represents reference input signal data at time (n−j).

The second adaptive filter 606 receives the second reference input signal q ₂ (n) and stores q ₂ (n) as buffer data bold b ₂ (n) in the signal buffer in the second adaptive filter 606. Above, the bold b ₂ (n) is filtered by the bold a ₂ (n) which is the second adaptive filter coefficient in accordance with the following equation (37) to output the second pseudo echo signal d ₂ (n).

The adder 607 receives d ₁ (n) and d ₂ (n), obtains a pseudo echo signal d (n) according to the following equation (38), and outputs it to the subtractor 610.

Next, in the example of FIG. 6, the filter set control unit 503 includes a buffer data converter 608 and a coefficient storage 609. The buffer data converter 608 converts the bold characters b ₁ (n) and bold characters b ₂ (n), which are buffer data stored in the signal buffer, at the transition of SEL (n). At the transition of SEL (n), the reference input signal output from the reference input signal generation unit 504 changes, so that the data bold characters b ₁ (n), bold characters b ₂ (n) stored in the signal buffer until then are newly There is a discrepancy between the input signal and the input signal.

That is, when SEL (n) = 1, the reference input signal is bold q ₁ (n) = bold l _in (n), bold q ₂ (n) = bold r _in (n), while SEL In the case of (n) = 2, bold q ₁ (n) = {bold l _in (n) + bold r _in (n)} / 2, bold q ₁ (n) = {bold l _in (n) −bold r _in (n)} / 2.

Therefore, the buffer data converter 608 converts the buffer data of the first adaptive filter 605 and the second adaptive filter 606 as follows when SEL (n) is switched. First, when changing from SEL (n−1) = 1 to SEL (n) = 2 from time (n−1) to time n, the buffer data converter 608 executes conversion according to the following equation (39). Conversely, when SEL (n-1) = 2 changes to SEL (n) = 1, conversion according to the following equation (40) is performed.

The coefficient storage 609 is a storage device for storing the coefficients of the adaptive filter, and has a storage buffer in which bold _letters a _bak1 (n) and bold _letters a _bak2 (n) are stored as buffer data. The coefficient storage 609 receives the filter selection result SEL (n), and when switching SEL (n), the bold a ₁ (n) that is the filter coefficient string of the first adaptive filter 605 and the second adaptive filter 606. Are replaced with the bold letter a ₂ (n).

That is, the contents of the bold a ₁ (n) in bold _a bak1 (n), stored in the storage buffer contents bold a ₂ (n) as bold _a bak2 (n), which up bold _a bak1 (n), The coefficient sequence stored in the storage buffer as bold a _bak2 (n) is _expanded into filter coefficient sequence bold a ₁ (n) and bold a ₂ (n).

  As described above, according to the fourth embodiment, the effect of reducing the size of the echo canceller can be obtained by reducing the number of adaptive filter sets that are operated simultaneously.

Embodiment 5. FIG.
In the fourth embodiment, when the adaptive filter set is replaced, the filter coefficient sequence of the unused adaptive filter set is backed up in the storage device, and this is alternately replaced according to the selection of the adaptive filter set. . On the other hand, in the fifth embodiment, the filter coefficient sequence is converted according to the relationship of the above equation (20).

  FIG. 7 is a block diagram showing a configuration of an echo canceling apparatus according to Embodiment 5 of the present invention. Components identical or corresponding to those in FIG. 6 are given the same reference numerals. The filter set control unit 503a according to the fifth embodiment includes a coefficient converter 611 in addition to the buffer data converter 608 described in the fourth embodiment. Other components and their functions are the same as those in the fourth embodiment, and the description thereof is omitted.

The coefficient converter 611 receives the filter selection result SEL (n), and converts the adaptive filter coefficients of the first adaptive filter 605 and the second adaptive filter 606 at the transition of SEL (n). More specifically, from the above equation (20), the transfer functions, bold letters h _L , bold letters h _R , bold letters h _S , bold letters h _D, have the relationship of the following formula (41).

As a result, the adaptive filter coefficient sequence of the adaptive filter that inputs l _in (n) and r _in (n) as reference signals is indicated by bold letters a _L (n) and bold letters a _R (n), and q _S (n) If the adaptive filter coefficient sequence of the adaptive filter set that inputs q _D (n) as a reference signal is indicated by bold a _S (n) and bold a _D (n), these relational expressions are as shown in the following formula (42). Become.

Therefore, the coefficient converter 611 performs the following filter coefficient conversion at the transition of SEL (n). That is, when SEL (n-1) = 1 changes from SEL (n-1) = 1 to SEL (n) = 2 from time (n-1) to time n, conversion according to the following equation (43) is performed, and SEL (n-2) = 1 When SEL (n) = 1, conversion according to the following formula (44) is performed.

  As described above, according to the fifth embodiment, when the adaptive filter set is replaced, the filter coefficient sequence is converted, and the result of adaptation learning up to the time of replacement is taken over by the next adaptive filter set. And the convergence of the adaptive filter can be accelerated.

Embodiment 6 FIG.
In the first embodiment, an example in which the adaptive filter set to be used is switched according to the degree of coincidence between the Left channel signal and the Right channel signal of the received signal has been described. In the sixth embodiment, the coefficient update step of the adaptive filter Only the gain is switched.

  FIG. 8 is a block diagram showing a configuration of an echo cancellation apparatus according to Embodiment 6 of the present invention. The echo cancellation apparatus according to the sixth embodiment includes a received signal analysis unit 801, a coefficient update step gain control unit 802, a first adaptive filter 803, a second adaptive filter 804, an adder 805, and a subtractor 806. .

The reception signal analysis unit 801 outputs the reception signal analysis result J (n) when the Left channel reception signal l _in (n) and the Right channel reception signal r _in (n) are input. When the analysis result J (n) is input, the coefficient update step gain control unit 802 outputs the coefficient update step gains α ₁ and α ₂ . The first adaptive filter 803 receives the Left channel reception signal l _in (n) and outputs the Left channel component pseudo echo signal d _L (n).

The second adaptive filter 804 receives r _in (n) and outputs a Right channel component pseudo echo signal d _R (n). The adder 805 inputs d ₁ (n) and d ₂ (n) and outputs a pseudo echo signal d (n). The subtractor 806 inputs d (n) and the transmission input signal s _in (n), and outputs a transmission output signal s _out (n).

The received signal analysis unit 801 has the same function as in the first embodiment, and a description thereof is omitted. The coefficient update step gain control unit 802 controls the coefficient update step gains α ₁ and α ₂ according to the received signal analysis result J (n). Here, α ₁ is a coefficient update step gain of the first adaptive filter 803, and α ₂ is a coefficient update step gain of the second adaptive filter 804.

Here, in the conventional echo canceller, the degree of coincidence between the bold channel l _in (n) that is the received signal time-series vector of the Left channel and the bold character r _in (n) that is the received signal time-series vector of the Right channel is high. That is, there is a problem that the adaptation error becomes large when a sound close to monaural is input.

Therefore, in the sixth embodiment, the coefficient update step gain control unit 802, based on the J (n), to determine the degree of matching bold l _in (n), bold _{r in (n), J (} n) If the value of is large, the received signal is close to monaural speech, so the values of α ₁ and α ₂ are lowered to reduce the adaptation error of the adaptive filter. Conversely, if the value of J (n) is small, α Increase the values of ₁ and α ₂ to accelerate the convergence of the adaptive filter.

If the coefficient update step gain when J (n) is sufficiently large is α _M1 and α _M2, and the coefficient update step gain when J (n) is sufficiently small is α _S1 and α _S2 , Each has the relationship of the following formula (45).

The values of α ₁ and α ₂ for J (n) are determined by threshold determination for J (n). Outputs α ₁ = α _M1, α ₂ = α _M2 if J (n) exceeds the threshold with respect to a predetermined threshold, and outputs α ₁ = α _S1, α ₂ = α _S2 if the threshold is below the threshold. To do. At this time, if the value of J (n) is near the threshold value of both, the values of α _{1 and} α ₂ may be adjusted stepwise according to the value of J (n).

When l _in (n) is input, the first adaptive filter 803 outputs a Left channel component pseudo echo signal d _L (n) according to the following equation (46). Further, when r _in (n) is input, the second adaptive filter 804 also outputs the Right channel component pseudo echo signal d _R (n) according to the following equation (46). d _L (n) and d _R (n) are added by the adder 805 as shown in the following equation (47) to output a pseudo echo signal d (n).

The subtractor 806 receives the pseudo echo signal d (n) and the transmission input signal s _in (n) and outputs the transmission output signal s _out (n). Using this s _out (n), the first adaptive filter 803 updates the bold a ₁ (n) that is the filter coefficient sequence, and the second adaptive filter 804 is the bold a that is the filter coefficient sequence. ₂ Update (n). As the coefficient update step gain at this time, α ₁ and α ₂ output from the coefficient update step gain control unit 802 are used. If the NLMS algorithm is used, the coefficient is updated according to the relationship of the following formula (48).

  As described above, according to the sixth embodiment, since only the coefficient update step gain of the adaptive filter is switched, the configuration is simpler than the configuration in which the adaptive filter set is replaced according to the received signal, and thus the apparatus is small-scale. The effect that it can be made is obtained.

It is a block diagram which shows the structure of the echo cancellation apparatus by Embodiment 1 of this invention. It is a block diagram which shows the structure of the adaptive filter set in FIG. It is a block diagram which shows the structure of the echo cancellation apparatus by Embodiment 2 of this invention. It is a figure which shows the structural example of the adaptive filter set in the echo cancellation apparatus by Embodiment 3 of this invention. It is a block diagram which shows the structure of the echo cancellation apparatus by Embodiment 4 of this invention. It is a figure which shows the structural example of the echo cancellation means in the echo cancellation apparatus by Embodiment 4 in detail. It is a block diagram which shows the structure of the echo cancellation apparatus by Embodiment 5 of this invention. It is a block diagram which shows the structure of the echo cancellation apparatus by Embodiment 6 of this invention. It is a figure explaining the recording method of a stereo sound.

Explanation of symbols

41, 101 first adaptive filter set, 42, 102 second adaptive filter set, 103, 501, 801 reception signal analysis unit (reception signal analysis unit), 104, 502 filter set selection unit (filter set selection unit), 105 First coefficient update control unit (first coefficient update control unit), 106 Second coefficient update control unit (second coefficient update control unit), 107, 108, 306, 506, 610, 806 Subtractor ( Subtracting means), 109 output signal switch (output signal switching means), 201, 303, 401, 605, 803 first adaptive filter, 202, 304, 402, 606, 804 second adaptive filter, 203, 305, 403, 408, 607, 805 Adder (adding means), 301, 404, 601 Sum signal generator (sum signal generating means), 302, 4 5,602 Difference signal generator (difference signal generation means), 205,406 Third adaptive filter, 407 Fourth adaptive filter, 503,503a Filter set control section (filter set control means), 504 Reference input signal generation section (Reference input signal generation means), 505 adaptive filter set, 603, 604 signal switcher (signal switch means), 608 buffer data converter (buffer data conversion means), 609 coefficient storage (coefficient storage means), 611 coefficient conversion (Coefficient conversion means), 802 coefficient update step gain control section (coefficient update step gain control means).

Claims (12)

Received signal analysis means for receiving and inputting two- channel audio signals and analyzing the degree of coincidence between the received signals of the respective channels;
A first adaptive filter set composed of adaptive filters corresponding to extraction of pseudo echo signals of the received signals of the two channels;
A second adaptive filter set including an adaptive filter corresponding to extraction of a pseudo echo signal of a monaural signal obtained from the received signals of the two channels;
A threshold is set for the degree of coincidence between the received signals of the respective channels, and the first adaptive filter set is selected when the degree of coincidence is lower than the threshold, and when the degree of coincidence is higher than the threshold Filter set selection means for selecting the second adaptive filter set ;
An echo canceller comprising echo canceling means for removing a pseudo echo signal from a transmission input signal using the adaptive filter set selected by the filter set selecting means and outputting as a transmission output signal. The echo cancellation means updates the adaptive filter coefficient when the first adaptive filter set is selected, and stops the update of the adaptive filter coefficient when the first adaptive filter set is not selected, and the second adaptive filter A second coefficient update control means for updating the adaptive filter coefficient when the set is selected, and stopping the update of the adaptive filter coefficient when the set is not selected; and a pseudo echo signal output from the first adaptive filter set. wherein the first subtracting means for excluding from the transmit input signal, a second subtracting means for excluding a pseudo echo signal from the transmission input signal output from the second adaptive filter set, the output from the first subtraction means to output the transmission input signal the pseudo echo signal is removed in the second of said selected adaptive filter set by switching the output from the subtraction means as a transmission output signal Echo canceller according to claim 1 Symbol mounting, characterized in that an output signal switching means. The first adaptive filter set receives a first adaptive filter that inputs one of the two-channel received signals and outputs a pseudo echo signal of the channel component, and inputs the other of the two-channel received signals. A second adaptive filter that outputs a pseudo echo signal of the channel component; and an adding means that adds the pseudo echo signals of both channel components and outputs the first pseudo echo signal;
The first adaptive filter and the second adaptive filter update an adaptive filter coefficient using a first residual signal obtained by removing the first pseudo echo signal from a transmission input signal by a first subtracting unit. ,
The second adaptive filter set includes sum signal generating means for inputting the received signals of the two channels to generate a sum signal, and a third adaptive filter for inputting the sum signal and generating a second pseudo echo signal. And
Said third adaptive filter claims, characterized in that updating the adaptive filter coefficients using the second residual signal except for the second pseudo echo signal from the transmit input signal by the second subtracting means 2. The echo canceller according to 2 . The first adaptive filter set receives a first adaptive filter that inputs one of the two-channel received signals and outputs a pseudo echo signal of the channel component, and inputs the other of the two-channel received signals. A second adaptive filter that outputs a pseudo echo signal of the channel component; and an adding means that adds the pseudo echo signals of both channel components and outputs the first pseudo echo signal;
The first adaptive filter and the second adaptive filter update an adaptive filter coefficient using a first residual signal obtained by removing the first pseudo echo signal from a transmission input signal by a first subtracting unit. ,
The second adaptive filter set includes a sum signal generating unit that inputs the two-channel received signal to generate a sum signal, a difference signal generating unit that inputs the two-channel received signal and generates a difference signal, A third adaptive filter that inputs the sum signal and outputs a pseudo echo signal of the sum signal component; a fourth adaptive filter that inputs the difference signal and outputs a pseudo echo signal of the difference signal component; and an adding means for outputting a second pseudo echo signal in addition to a pseudo echo signal of the difference signal component and the signal component,
The third adaptive filter and the fourth adaptive filter update the adaptive filter coefficient using the second residual signal obtained by removing the second pseudo echo signal from the transmission input signal by the second subtracting means. The echo canceller according to claim 2 . Received signal analysis means for receiving and inputting two- channel audio signals and analyzing the degree of coincidence between the received signals of the respective channels;
An adaptive filter set composed of an adaptive filter that extracts a pseudo echo signal of the input signal;
A threshold is set for the degree of coincidence between the received signals of each channel, and when the degree of coincidence is lower than the threshold, an adaptive filter set that matches the two-channel signal is selected, and the degree of coincidence is higher than the threshold In this case, filter set selection means for selecting an adaptive filter set that matches the monaural signal obtained from the received signals of the two channels ,
Reference input signal generation means for outputting the two- channel signal or the monaural signal to the adaptive filter set according to the selection result of the adaptive filter set;
Filter set control means for setting an adaptive filter coefficient adapted to the two- channel signal or an adaptive filter coefficient adapted to the monaural signal in the adaptive filter set according to a selection result of the adaptive filter set;
An echo canceller comprising echo canceling means for removing a pseudo echo signal from a transmission input signal using the adaptive filter set selected by the filter set selecting means and outputting as a transmission output signal. Reference input signal generating means includes a sum signal generating means for generating a sum signal of two channels of a received signal, and a difference signal generating means for generating a difference signal of the received signal of the two channels, compatible with 2-channel signal adaptive filter When a set is selected, the received signals of the two channels are output as reference input signals to the adaptive filter set, and when an adaptive filter set that matches a monaural signal is selected, the sum signal and the difference signal The echo canceling apparatus according to claim 5, further comprising: a signal switching unit that outputs the signal as a reference input signal to the adaptive filter set. When the selection of the adaptive filter set is changed , the filter set control means refers to the time series data of the past reference input signal held in the data buffer in the adaptive filter set that matches the selected adaptive filter set . Buffer data conversion means for converting to an input signal value and the adaptive filter coefficient of the adaptive filter set to be changed according to the selection result of the adaptive filter set, and corresponding to the selected adaptive filter set that has been saved 7. An echo canceller according to claim 6, further comprising coefficient storage means for setting an adaptive filter coefficient to be used. When the selection of the adaptive filter set is changed , the filter set control means refers to the time series data of the past reference input signal held in the data buffer in the adaptive filter set that matches the selected adaptive filter set . a buffer data conversion means for converting the input signal value, the adaptive the filter set Ru is changed, the coefficient conversion means for converting to conform to the newly selected adaptive filter set adaptive filter coefficients of the adaptive filter set before the change DOO echo canceller according to claim 6 Symbol mounting characterized by comprising a. Received signal analysis means for receiving and inputting two- channel audio signals and analyzing the degree of coincidence between the received signals of the respective channels;
A threshold is set for the degree of coincidence between the received signals of each channel, the adaptive update step gain is decreased when the degree of coincidence is higher than the threshold, and the coefficient is updated when the degree of coincidence is lower than the threshold. Coefficient update step gain control means for controlling to increase the step gain ;
A plurality of adaptive filters that input the reception signals of the two channels to generate a pseudo echo signal for each channel component and update adaptive filter coefficients using the transmission output signal ;
Adding means for adding a pseudo echo signal of each channel component and outputting as a pseudo echo signal;
Echo canceller which includes a subtraction means for a signal excluding the pseudo echo signal the adding means is output from the transmission input signal, and outputs it as the transmit output signal. Received signal analyzing means any one of claims 9 to evaluate the matching degree on the basis of the signal power ratio between the sum signal and the difference signal between the two channels of a received signal from claim 1, wherein The echo canceller described. The echo cancellation apparatus according to any one of claims 1 to 9 , wherein the reception signal analysis means evaluates the degree of coincidence based on a cross-correlation between the reception signals of the two channels. Sum signal generating means for generating a sum signal between the received signals of the two channels;
Difference signal generating means for generating a difference signal between the two- channel signals;
A first adaptive filter that inputs the sum signal output from the sum signal generation means and outputs a pseudo echo signal of the sum signal component and updates the adaptive filter coefficient using the transmission output signal ;
A second adaptive filter that inputs the difference signal output from the difference signal generation means and outputs a pseudo echo signal of a difference signal component, and updates an adaptive filter coefficient using the transmission output signal ;
Adding means for outputting a pseudo echo signal obtained by adding the pseudo echo signal of the sum signal component and the pseudo echo signal of the difference signal component;
Echo canceller which includes a subtraction means for a signal obtained by subtracting the pseudo echo signal the adding means is output from the transmission input signal, and outputs it as the transmit output signal.

Images