JP3202256B2 - Acoustic echo canceller for digital voice communication equipment. - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an acoustic echo canceller provided for canceling an acoustic echo generated during a hands-free call in a voice communication device having a hands-free call mode, such as an automobile radio telephone device. .

[0002]

2. Description of the Related Art Conventionally, some radio telephones for vehicles such as automobile radio telephones have a hands-free telephone call mode in addition to a handset telephone call mode. Hands-free call mode is a mode in which a speaker for receiving and a microphone for transmitting are installed on a communication device or dashboard etc. separately from the handset, and these speakers and the microphone can be used as a handset to make a call. It was made. With this hands-free call mode, the talker can talk without having to hold the handset. Is remarkably effective.

[0003] However, when such a hands-free call mode is used, a received voice output from a speaker goes around the microphone during a call, and is transmitted to the other party to generate an acoustic echo. For this reason, this type of telephone device is generally provided with an acoustic echo canceller for canceling the acoustic echo. Such an acoustic echo canceller is indispensable for a radio telephone apparatus for digitizing and transmitting an audio signal, especially since a signal delay due to a codec or the like is large.

FIG. 4 shows an example of the configuration of this type of echo canceller together with peripheral circuits. In the figure, when an acoustic echo path EC exists between the speaker 12 and the microphone 13, a part of the received voice output from the speaker 12 becomes an acoustic echo and
3 is input. Echo signal ES by the above acoustic echo
Is converted into a digital signal by the A / D converter 14, and then input to the acoustic echo canceller 100. In the acoustic echo canceller 100, the echo signal ES is input to an adaptive filter 101 as a residual signal TS via a subtractor 102. The adaptive filter 101 performs an adaptive operation for reducing the residual signal TS to zero. The adaptive operation estimates the acoustic characteristics of the used space, that is, the echo path EC, and the estimation result and the received signal RS
, A pseudo echo signal EES is generated. Subtractor 10
2, the pseudo echo signal EES is obtained from the echo signal ES.
, And the echo signal ES is cancelled.

By the way, the adaptive operation of the adaptive filter 101 is normally performed in a single talk state in which only a received signal exists, but during reception, for example, a voice of a surrounding person is input to a microphone. When the double talk state occurs, the acoustic echo canceller 100 performs an adaptive operation to cancel not only the echo signal ES but also the voices of the surrounding people, and as a result, the estimation result of the echo path EC is confused.

Therefore, conventionally, for example, a double talk detector 103 is provided as shown in FIG. 4, and the level ratio R between the residual signal TS and the received signal RS is determined by this detector 103 as R = log
The calculation is performed as {(signal level of reception signal RS) / (signal level of residual signal TS)}, and this level ratio R is compared with a preset threshold value Dth (k). When the level ratio R becomes smaller than the threshold value Dth (k), it is determined that a double talk state has occurred, and during the period in which the double talk state is detected, the adaptive filter 1
01 is stopped. In this way, it is possible to prevent the acoustic echo canceller 100 from erroneously estimating the echo path EC in the state of double talk, and to realize a stable echo canceling operation.

Recently, in order to improve the detection accuracy and the detection speed of the double talk state, the threshold value Dth (k) is changed according to the detection result of the Bull Talk state at the immediately preceding detection timing. Is considered. That is, in the single talk state, the threshold value Dth (k + 1) is set to Dth (k + 1) = Dth (k) + δ1 to make it easier to detect the start timing of the double talk. During the period in which the state is detected, the threshold value Dth (k + 1) is set to Dth (k + 1) = Dth (k) -δ2 in order to easily detect the end point of the double talk.

[0008]

However, such an acoustic echo canceller 101 has the following problems to be improved. That is, as described above, the threshold value Dth
(k + 1) is a high value (Dth (k)
+ .Delta.1), and is unconditionally changed to a low value (Dth (k)-. Delta.2) when a double talk state is detected.
For this reason, although a double talk state has occurred, for example, the signal level of the residual signal TS is not so large as compared with the signal level of the reception signal RS, and the level ratio R is (Dth (k) + δ1) ≧ level R ≧ ( Dth (k)-. Delta.2), the detection result alternately changes between single talk (wrong decision) and double talk (correct decision).

In such a state, the adaptive filter intermittently performs the adaptive operation in spite of the double talk state, so that the estimation result of the echo path changes unnecessarily, and as a result, The cancel operation becomes extremely unstable. In addition, unnecessary operation of the adaptive filter wastes power, thereby shortening the battery life of the device.

Therefore, the present invention shortens the period in which the detection result of double talk falls alternately between correct judgment and erroneous judgment, thereby improving the stability of double talk detection and, consequently, the stability of echo canceling operation. It is an object of the present invention to provide an acoustic echo canceller of a voice communication device capable of improving the acoustic echo.

[0011]

In order to achieve the above object, the present invention employs the following means. That is, it is well known that one of the main causes of double talk is transmission delay of a speech signal. In a digital voice communication device, this transmission delay mainly occurs due to the processing time of various codecs such as a voice codec.
It cannot be avoided unless a telephone call is made using a digital voice communication device. In other words, once in the double talk state, the double talk state is continuous at least for a period corresponding to the transmission delay, and it is not necessary to detect the end of the double talk in this period.

The present invention has been made in view of this point, and detects an echo canceller main body, a double talk state, and stops the echo path estimating operation of the echo canceller main body during the double talk state detection period. Threshold control means is provided in addition to double talk detection means. The threshold control means sets the threshold to a value at which the double talk state can be easily detected during the period when the double talk state is not detected by the double talk detecting means, while the double talk state is detected. In the period during which the threshold value is set, the threshold value is maintained at a value that makes it easy to detect the double-talk state for a predetermined period from the time of detection of the double-talk state, that is, a period corresponding to the transmission delay described above. The threshold is changed and set to a value that makes it easier to detect the end of the double talk state.

As a result, according to the present invention, the threshold value is unconditionally held at a value that makes it easy to detect the double talk state, at least until a period corresponding to the transmission delay has elapsed even when the double talk state is entered. You. Therefore, even if the signal level of the residual signal is not so large as compared to the signal level of the received signal, during the period corresponding to the transmission delay,
The problem that the detection result of double talk alternately changes between single talk (false determination) and double talk (correct determination) is prevented. Therefore, the period during which the double-talk detection result becomes unstable is shortened, which makes it possible to increase the stability of the echo cancellation operation compared to the case where the threshold value is unconditionally changed according to the double-talk detection result. become. Further, it is possible to reduce unnecessary power consumption by the echo canceller. further,
When the period corresponding to the transmission delay has elapsed, the threshold value is always changed to a value for detecting the end of double talk. Therefore, it is possible to reliably detect the transition from the double talk state to the single talk state, whereby the adaptive operation of the adaptive filter can be started quickly.
Further, according to the present invention, in the period in which the double talk state is detected, the threshold value is further set to be greater than the first value in the period corresponding to the transmission delay of the speech signal from the time when the double talk is detected. The end value is changed to a third value that is difficult to detect, and is maintained at this value. In the subsequent period, the threshold value is changed to a second value that is easier to detect the end of the double talk state than the first value. It is also characterized. By controlling in this way, during the period corresponding to the transmission delay,
In spite of the double talk state, it is possible to further reduce the problem that single talk is erroneously detected. That is, the detection state of double talk can be held more reliably.

[0014]

FIG. 2 is a circuit block diagram showing a configuration of an automobile radio telephone apparatus provided with an acoustic echo canceller according to an embodiment of the present invention. This apparatus is a so-called dual mode type apparatus using both a digital mode and an analog mode.

In FIG. 1, a radio carrier signal arriving via a radio line is received by an antenna 1 and then input to a receiving circuit (RX) 3 via a duplexer (DUP) 2. In the receiving circuit 3, the received carrier signal is combined with the local oscillation signal output from the frequency synthesizer (SYN) 4 and converted into an intermediate frequency signal. The received intermediate frequency signal is input to a digital demodulation circuit (DEM) 5 in the digital mode, and is input to an analog signal processing circuit (ANALOG A) in the analog mode.
(UDIO) 20.

First, the case where the digital mode is set will be described. In this case, the changeover switches 6, 16
Are switched to the digital demodulation circuit 5 and the digital modulation circuit (MOD) 17 according to the switching control signal S1 output from the control circuit 31. In the digital demodulation circuit 5, the received intermediate frequency signal is digitally demodulated and converted into an analog baseband signal. This baseband signal is converted into a digital signal by an A / D converter 7 and then converted into a channel codec (CH-CODEC) 8.
, An error correction process is performed, and then a decoding process is performed by a speech codec (SPEACH-CODEC) 9.
The digital reception signal output from the speech codec 9 is converted into an acoustic echo canceller (ECHO-
D / A converter 11 after passing through CANCELLER 10
Is converted into an analog signal, and is output from the speaker 12 in a loud sound.

On the other hand, the transmission signal output from the microphone 13 is converted into a digital signal by the A / D converter 14 and then input to the acoustic echo canceller 10. In the acoustic echo canceller 10, a process for canceling an echo signal generated when a part of the received voice goes around the microphone 13 is performed. The transmission signal on which the echo cancellation processing has been performed is encoded by the speech codec 9 and then subjected to error correction encoding processing by the channel codec 8. Then, the digital transmission signal is converted into an analog signal by the D / A converter 15 and then input to the digital modulation circuit (MOD) 17 via the changeover switch 16. In the digital modulation circuit 17, digital modulation of a carrier is performed using the digital transmission signal as a modulation signal. As the digital modulation method, for example, a QPSK method is used. The transmission signal output from the digital modulation circuit 17 is combined with a local oscillation signal output from the frequency synthesizer 4 by a transmission circuit (TX) 18 and converted into a signal in a radio frequency band. Transmitted from the antenna 1 via the antenna 1.

When the analog mode is set, the analog signal processing circuit 20 is inserted into the transmission / reception channel by the changeover switches 6 and 16. The circuit 20 performs band limitation and expansion processing on the received intermediate frequency signal, and performs necessary signal processing such as band limitation and compression processing on the transmission signal.

In the figure, reference numeral 40 denotes an outgoing switch, a dial key,
Console unit (CU) with display, etc., 5
Reference numeral 0 denotes a power supply circuit for generating a predetermined power supply voltage Vcc based on the output of the vehicle battery 51.

The acoustic echo canceller 10
Is configured as follows. FIG. 1 is a circuit block diagram showing the configuration. In this figure, the same parts as those in FIG. 4 are denoted by the same reference numerals. The acoustic echo canceller 10 includes an adaptive filter 101, a subtractor 102, and a double talk detection circuit 110.

Of these, the double talk detection circuit 110
It has absolute value detectors (ABS) 111 and 115,
These absolute value detectors 111 and 115 detect the absolute values of the amplitudes of the reception signal RS and the residual signal TS, respectively. At this time, the past N data stored in the registers 118 and 119 are also input to the absolute value detectors 111 and 115. The absolute values of the received signal RS and the residual signal TS detected by the absolute value detectors 111 and 115 are input to peak detectors 112 and 116, respectively, where the peak values are detected. Then, these peak detection values are logarithmically converted by logarithmic converters 113 and 117, respectively, and then input to a subtractor 114, where the difference is detected. That is, the level ratio R between the peak value X of the received signal RS and the peak value E of the residual signal ES, that is, logX-logE = R is obtained by the above processing.

The calculated level ratio R is compared with the threshold value Dth output from the threshold value control section 121 in the comparator 120. When the level ratio R ≦ Dth, the comparator 120 outputs A detection signal indicating that a double talk state has occurred is output. The filter coefficient update unit 123 prohibits the update of the filter coefficient of the adaptive filter 101 during the period in which the double talk state is detected, according to the detection signal output from the comparator 120.

On the other hand, the detection signal output from the comparator 120 is supplied to a threshold control unit 121 and a counter 122.
Is also entered. The counter 122 starts counting from the time when the detection signal is generated, and outputs a time-over signal when a predetermined time has elapsed.

The threshold control unit 121 is provided with the comparator 12
Based on the comparison result of 0 and the output of the counter 122, the threshold value of the comparator 120 is variably controlled as follows.
That is, in a period in which double talk is not detected, that is, in a single talk state, a relatively high threshold value Dth (k + 1) = Dth (k) + δ1 for easily detecting double talk is output. On the other hand, when the double talk state is detected, the high threshold is held for a period from the time of detection to the time when the counter 122 times out, and after the time over, the end of the double talk is detected until the end of the double talk is detected. During the period, this threshold value is output after changing to a relatively low value Dth (k + 1) = Dth (k)-. Delta.2 which makes it easy to detect the end of the time-over.

The time counted by the counter 122 is set as follows, for example. That is, in a state where the speaker A and the speaker B are talking, for example, the speaker B
It is assumed that the speaker A starts speaking during the transmission, and the speaker B receives it and immediately stops the transmission. In this case, in the apparatus of the speaker B, the transmission signal BTS and the reception signal BRS coming from the speaker A do not exist at the same time as shown in FIG. That is,
Does not enter double talk state. However, the reception signal BRS in the apparatus of the speaker B is the transmission signal ATS transmitted earlier by the transmission delay amount T1 in the apparatus of the speaker A, and the transmission signal BTS of the speaker B is the transmission delay T2. Later, the device of speaker A receives the received signal ARS. Therefore, in the apparatus of the speaker A, a double talk state always occurs due to the transmission delays T1 and T2 as shown by the hatched portion in FIG. That is, the double talk state continues in any case for at least a period corresponding to the total value T of the transmission delays T1 and T2. In other words, this means that it is not necessary to detect the end of the double talk at least during the period corresponding to the total value T of the transmission delay in the period of the double talk state. Therefore, paying attention to this point, the total value T of the transmission delay is set in the counter 122 as the clocking time.

Next, the operation of the acoustic echo canceller 10 configured as described above will be described. First, in a period in which double talk is not detected, that is, in a single talk state, the threshold control unit 121 outputs a high threshold value (Dth (k) + δ1) at which double talk is easily detected. Therefore, the comparator 120 monitors the double talk using the high threshold value.

Now, double talk occurs in this state,
As a result, it is assumed that the level ratio R between the received signal and the residual signal is reduced. Then, this decrease in the level ratio R is immediately detected by the high threshold in the comparator 120,
The result is notified to the filter coefficient changing unit 123. For this reason, the filter coefficient changing unit 123 performs a process for inhibiting the coefficient change of the adaptive filter 101 until the end of the double talk is detected. Therefore,
In the subsequent double talk state, the adaptive filter 10
Even if the adaptive operation is performed in step 1 according to a signal other than the echo signal ES, the estimation result of the echo path EC is not confused.

When a double talk state is detected in the comparator 120, the threshold value control unit 121 does not immediately change the threshold value to a low value at which the end of double talk can be easily detected. A high value for double talk detection is maintained for a predetermined period T corresponding to the amount. When the counter 122 outputs a time-over signal after the lapse of the predetermined time T, the threshold control unit 121 sets the threshold to a low value (Dth (k)) at which the end of double talk is easily detected.
−δ2). Therefore, the residual signal TS
Is not so large as compared with the signal level of the reception signal RS, and even if the level ratio R does not sufficiently decrease, the detection result of the double talk during the above-mentioned predetermined period T is a single talk (erroneous determination) and a double talk. There is no problem that the change is repeated alternately with (correct judgment).

After the predetermined period T has elapsed,
As described above, the threshold value is immediately changed to a low value (Dth (k)-. Delta.2) at which the end of double talk is easily detected. Therefore, even if the double talk state ends immediately after the elapse of the predetermined period T, the end of the double talk is quickly and reliably detected by the low threshold value.

As described above, according to the present embodiment, the threshold value is set by focusing on the fact that the double talk state is continued only during the predetermined period T corresponding to the transmission delay amount from the time when the double talk state is detected. Because the signal level of the residual signal TS is not so large as compared with the signal level of the reception signal RS and the level ratio R does not sufficiently decrease, the detection result of the double talk is unstable because the value is held at a high value. Don't worry. Therefore, the stability of the echo cancellation operation can be improved. Further, unnecessary adaptive operation is not performed by the adaptive filter 101, thereby reducing unnecessary power consumption by the echo canceller. This effect is extremely effective for extending the battery life in a wireless telephone device using a battery as a power supply.

The present invention is not limited to the above embodiment. For example, in the above embodiment, the threshold is not changed in the predetermined period T corresponding to the transmission delay amount from the time when the double talk state is detected, but the threshold is changed to a value higher than this threshold. It may be.
By doing so, it is possible to further reduce the problem that single talk is detected in spite of double talk during the predetermined period T. That is, the detection state of double talk can be held more reliably.

In a predetermined period T corresponding to the amount of transmission delay, the comparator 1
20. The comparison operation 20 is stopped, and the value of the level ratio R input to the comparator 120 is latched to the value at the time when double talk is detected, or held at an ideal value in a preset double talk state. You may do so.

As a means for measuring the predetermined period T, a monostable multivibrator may be used in addition to the counter 122, and the configuration, control procedure and control contents of the threshold control means, and the double talk detecting means may be used. Regarding the circuit configuration, the type and configuration of the voice communication device,
Various modifications can be made without departing from the spirit of the present invention.

As described in detail above, according to the present invention, the threshold value is set to the first value in which the double talk state can be easily detected during the period when the double talk state is not detected by the double talk detecting means.
During the period during which the double talk state is detected, the threshold value is held at the first value during a period corresponding to the transmission delay of the speech signal from the time when the double talk is detected. In the subsequent period, the threshold value is changed to a second value that makes it easier to detect the end of the double talk state. Therefore, according to the present invention, during the period corresponding to the transmission delay, the detection result of the double talk is prevented from falling into a state in which the double talk alternately changes between the single talk (erroneous determination) and the double talk (correct determination). Accordingly, it is possible to provide an acoustic echo canceller of a digital voice communication device capable of improving the stability of double talk detection and the stability of an echo canceling operation. Further, according to the present invention, in the period in which the double talk state is detected, the threshold value is set to be larger than the first value in the period corresponding to the transmission delay of the speech signal from the time when the double talk is detected. The end of the state is changed to a third value that is easy to detect, and is maintained at this value. In the subsequent period, the threshold value is set to a second value that makes it easier to detect the end of the double talk state than the first value. With the change, it is possible to further reduce the problem that single talk is erroneously detected in the period corresponding to the transmission delay despite the double talk state. That is,
It is possible to provide an acoustic echo canceller of a digital voice communication device that can more reliably hold a double talk detection state.

[Brief description of the drawings]

FIG. 1 is a circuit block diagram showing a configuration of a dual mode type automobile radio telephone apparatus provided with an acoustic echo canceller according to an embodiment of the present invention.

FIG. 2 is a circuit block diagram showing a configuration of an acoustic echo canceller according to one embodiment of the present invention.

FIG. 3 is a signal waveform diagram for explaining a state of occurrence of double talk.

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

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Antenna, 2 ... Duplexer, 3 ... Receiving circuit, 4 ... Frequency synthesizer, 5 ... Digital demodulation circuit, 6, 16 ... Changeover switch, 7, 14 ... A / D converter, 8 ... Channel codec, 9 ... Speech Codec, 10 acoustic echo canceller, 11, 15 D / A converter, 12 speaker, 13 microphone, 17 digital modulation circuit,
18 ... transmission circuit, 20 ... analog signal processing circuit, 30 ...
Control circuit, 40: console unit, 50: power supply circuit, 51: battery, 101: adaptive filter, 102
... subtracter, 110 ... double talk detection circuit, 111, 1
15: absolute value detector, 112, 116: peak detector,
113, 117 ... logarithmic converter, 114 ... subtractor, 11
8, 119 register, 120 comparator, 121 threshold control unit, 122 counter, EC echo path, E
S: echo signal, RS: reception signal, TS: residual signal (transmission signal).

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H04B 3/23 H04M 1/60

Claims (2)

(57) [Claims] An echo path in a speech space is estimated from an echo signal generated when a part of a reception sound output from a speaker goes around a microphone during a call, and a pseudo echo is generated based on the estimation result and the reception signal. A signal is generated, and an echo canceller body for canceling the echo signal by subtracting the pseudo echo signal from the echo signal, and a level ratio between a reception signal and a speech signal passing through the echo canceller body is compared with a threshold value. A double talk detecting means for detecting a double talk state, and stopping an echo path estimation operation of the echo canceller main body during the detection period of the double talk state; and a double talk state is not detected by the double talk detecting means. period easy to detect the threshold of the double-talk state to a first value Set, whereas among the periods double talk state is detected, the Daburuto
From the point at which the
Threshold value control means for holding the threshold value at the first value during a period, and changing the threshold value to a second value at which the end of the double talk state is easily detected in a period thereafter. An acoustic echo canceller for a digital voice communication device. 2. An echo path in a communication space is estimated from an echo signal generated when a part of a reception sound output from a speaker goes around a microphone during a call, and a pseudo echo is generated based on the estimation result and the reception signal. A signal is generated, and an echo canceller body for canceling the echo signal by subtracting the pseudo echo signal from the echo signal, and a level ratio between a reception signal and a speech signal passing through the echo canceller body is compared with a threshold value. A double talk detecting means for detecting a double talk state, and stopping an echo path estimation operation of the echo canceller main body during the detection period of the double talk state; and a double talk state is not detected by the double talk detecting means. In the period, the threshold is set to a first value at which the detection of the double talk state is easy. On the other hand, during the period in which the double talk state is detected, the threshold value is further set to be greater than the first value in the period corresponding to the transmission delay of the speech signal from the time when the double talk is detected. Is changed to a third value which is hard to detect the end of the state, and is kept at this value. In the subsequent period, the threshold value is set to a second value which makes it easier to detect the end of the double talk state than the first value. An acoustic echo canceller for a digital voice communication device, comprising: threshold value changing means for changing.