JP4743085B2 - Echo canceller - Google Patents

Description

  The present invention relates to an echo canceller and an echo canceller program, and can be applied to, for example, an IP telephone (including a soft phone) that performs voice communication via an IP network.

  The echo canceller estimates the state of an echo path in which the received signal wraps around the transmitted signal, creates a pseudo echo signal from the received signal, and removes the echo component in the transmitted signal by canceling it with the pseudo echo signal.

  Here, in the case of an existing telephone, the delay time of the received signal in the echo path of the acoustic echo that goes from the microphone to the speaker or the line echo in the hybrid circuit varies depending on the temperature, but the variation is small. Therefore, even if the pseudo echo signal creation unit that creates the pseudo echo signal (such as the adaptive filter and the learning configuration of the filter coefficients) starts the learning algorithm for estimating the echo path state, the initial value of the delay time is set as a fixed value. No problem.

  Today, IP telephones, including soft phones, are beginning to become quite popular. In such IP telephones and the like, from when the received signal is output from the received signal output terminal of the echo canceller, until the echo component related to the received signal returns to the transmitted signal input terminal of the echo canceller Some delay times vary significantly over small changes such as due to temperature changes. For example, the processing time and processing timing of the received signal processing system (for example, processing waiting buffer) after leaving the echo canceller, or until the transmission signal captured by the microphone is input to the echo canceller. The delay time varies greatly due to variations in processing time and processing timing by the processing system. In this way, although the delay time variation of the echo component is large, if the initial value of the delay time when starting the learning algorithm is fixed, the actual delay time and the initial value of the delay time are greatly different. As a result, the convergence during learning of the echo path state becomes slow, and the echo component cannot be effectively removed.

For this reason, it is conceivable to detect the delay time and set it to the initial value of the delay time when starting the learning algorithm. It is also conceivable to capture the delay time by correlation processing between the reception signal output from the echo canceller and the transmission signal output to the echo canceller (see, for example, Non-Patent Document 1).
Naoki Mikami, “Introduction to Digital Signal Processing-Basics and Applications of Waveform Processing Technology in the Microcomputer Era”, p. 107, published by CQ Publisher 1988

  When trying to advance the timing at which the pseudo echo signal can be generated appropriately, it is desirable to detect the delay time value set in the learning algorithm even in the double talk state.

  However, if the far-end speaker related to the received signal and the near-end speaker related to the transmitted signal are similar and both use the same expression in the call (speak the same term because it is a conversation) The possibility is high), and not because of the correlation between the received signal and the echo component, the correlation may be high, and the delay time may be erroneously detected. If an incorrect delay time is set to a setting value used in the learning algorithm, naturally, convergence in learning is also delayed.

  Note that the above-described problem occurs not only when the echo canceller is realized by hardware, but also when the echo canceller is realized by the CPU executing an echo canceller program.

  Therefore, an echo canceller and an echo canceller program that can detect the delay time related to the echo path with high accuracy and increase the learning accuracy of the echo path state are desired.

  According to a first aspect of the present invention, there is provided an echo canceller for removing an echo component included in a transmission signal by using a pseudo echo signal created by a pseudo echo signal creating means having an adaptive filter and a filter coefficient updating unit. A reception signal filter for generating a signal limited to a band having a large individual difference in the voice signal from the reception signal input to the echo canceller; and (2) an individual in the voice signal from the transmission signal input to the echo canceller. A transmission signal filter for generating a signal limited to a band having a large difference; and (3) searching for a time difference at which the correlation between the signal generated by the reception signal filter and the signal generated by the transmission signal filter is high. And delay time detecting means for detecting a delay time in the echo path and supplying the delay time to the filter coefficient updating unit.

  The echo canceller program according to the second aspect of the present invention includes a computer, (1) a pseudo echo signal creating means having an adaptive filter and a filter coefficient updating unit for creating a pseudo echo signal, and (2) input to the echo canceller. A reception signal filter for generating a signal limited to a band having a large individual difference in the voice signal from the reception signal, and (3) a transmission signal input to the echo canceller is limited to a band having a large individual difference in the voice signal. A transmission signal filter for generating a signal; and (4) searching for a time difference in which the correlation between the signal generated by the reception signal filter and the signal generated by the transmission signal filter is high, and calculating a delay time in the echo path. It is characterized by functioning as a delay time detecting means for detecting and giving to the filter coefficient updating unit.

  According to the present invention, it is possible to provide an echo canceller and an echo canceller program that can detect the delay time related to the echo path with high accuracy and increase the learning accuracy of the echo path state.

(A) Main Embodiment Hereinafter, an embodiment of an echo canceller and an echo canceller program according to the present invention will be described in detail with reference to the drawings.

(A-1) Configuration of Embodiment FIG. 1 is a block diagram showing an internal configuration of a telephone terminal including an echo canceller (echo canceller unit) of the embodiment.

  In FIG. 1, a telephone terminal 100 exchanges call signals (received signal and transmitted signal) with other telephone terminals (not shown) via a network NW. The network NW is arbitrary, but is an IP network, for example.

  The telephone terminal 100 includes a network communication unit 101, an echo canceller unit 102, a speaker unit 103, and a microphone unit 104 (the operation unit and the display unit are omitted).

  The network communication unit 101 performs communication with the network NW. For example, the network communication unit 101 performs packet assembly, disassembly, channel modulation, demodulation, and the like.

  The speaker unit 103 outputs a received signal as a sound. In the case of FIG. 1, the speaker unit 103 appropriately includes functions such as rate conversion, buffering, and digital / analog conversion.

  The microphone unit 104 captures the transmission of a near-end speaker. In the case of FIG. 1, the microphone unit 104 appropriately includes functions such as a filter, analog / digital conversion, and buffering.

  For example, in this embodiment, a wideband audio signal of 8 kHz or less can be communicated, and the sampling frequency in the digital signal is 16 kHz.

  The echo canceller unit 102 removes an echo component of the received signal from the transmitted signal. In the case of this embodiment, for example, the echo canceller unit 102 is configured by the installed echo canceller program being executed by the CPU of the telephone terminal 100, but functionally, FIG. As shown, the reception signal filter 200, the transmission signal filter 201, the delay calculation unit 202, the pseudo echo creation unit 203, and the echo cancellation unit 204 are included.

  Note that some functions of the network communication unit 101 may be prepared as a program integrated with the echo canceller program.

  The reception signal filter 200 generates a signal limited to a band having a large individual difference from the reception signal. As the reception signal filter 200, for example, a band-pass filter that passes a component of 2 kHz to 4 kHz, a high-pass filter that passes a component of 2 kHz or more (may be a low-pass removal filter that excludes a component of 2 kHz or less), and the like. Can be applied.

  The transmission signal filter 201 generates a signal limited to a band having a large individual difference from the transmission signal before the echo component is removed. As the transmission signal filter 201, a band pass filter or a high pass filter similar to the reception signal filter 200 can be applied.

  The same filter may be used as the reception signal filter 200 or the transmission signal filter 201 in a time division manner.

  FIG. 2 is an explanatory diagram conceptually showing the frequency characteristics of the voice signals of a plurality of people. As shown in FIG. 2, the frequency characteristics of persons A, B, and C are the same for components of 2 kHz or less, but individual differences are large for components of 2 kHz or more. By using such a band with a large individual difference, it is easy to discriminate speakers. Speaking only from the frequency characteristic of FIG. 2, it is sufficient to use a component of 2 kHz or more, but a band of 4 kHz or more is easily affected by noise. For this reason, a bandpass filter that allows the components of 2 kHz to 4 kHz to pass as described above is preferable.

  The delay calculation unit 202 obtains a delay time in the echo path by searching for a time difference having the highest correlation between the signal output from the reception signal filter 200 and the signal output from the transmission signal filter 201. . That is, the delay calculation unit 202 calculates a delay time until the reception signal that has been output from the echo canceller unit 102 returns to the transmission signal to the echo canceller unit 102 as an echo component.

  The pseudo echo creating unit 203 includes an adaptive filter and a filter coefficient updating unit, and creates a pseudo echo signal from a received signal (received signal not filtered by the received signal filter 200) or the like. The delay time calculated by the delay calculation unit 202 is used as the initial value of the delay time required when the filter coefficient update unit of the pseudo echo creation unit 203 starts or restarts the update algorithm. As an algorithm for updating the filter coefficient, an algorithm according to an existing LMS method, an affine projection method, or other methods can be applied.

  The echo cancellation unit 204 removes an echo component included in the transmission signal from the microphone unit 104 by using the pseudo echo signal from the pseudo echo generation unit 203, and the transmission signal after the removal is networked. This is given to the communication unit 101.

(A-2) Operation of Embodiment Next, the operation of the echo canceller (echo canceller unit 102) of the embodiment having the above-described configuration will be described together with the peripheral operations. FIG. 3 is a flowchart showing a schematic processing flow when the echo canceller unit 102 is configured around an echo canceller program. However, in practice, a plurality of processes are processed in parallel.

  When the communication path with the opposite terminal is established by the network communication unit 101 (step 301), the echo canceller unit 102 acquires the received signal and outputs the sound from the speaker unit 104 (steps 302 and 303). The transmission signal captured by the microphone unit 104 is input to the echo canceller unit 102 (step 304).

  The reception signal output from the echo canceller unit 102 and the transmission signal input to the echo canceller unit 102 are respectively input to the corresponding reception signal filter 200 or transmission signal filter 201. The signal passes through the reception signal filter 200 or the transmission signal filter 201 and is given to the delay calculation unit 202 (step 305). Based on the correlation between the two signals filtered only in the band having a large individual difference, the delay calculation unit 202 calculates the delay time in the echo path and provides it to the pseudo echo creation unit 203 (step 306).

  In the pseudo echo creation unit 203, a filter coefficient update algorithm is started with a given delay time as an initial value, and a pseudo echo signal is generated by adaptive filter processing using the filter coefficient at each time point (step 307). ). The echo canceling unit 204 cancels the echo component included in the transmission signal from the microphone unit 104 with the pseudo echo signal from the pseudo echo creating unit 203 and removes the operation signal after the removal. The communication unit 101 transmits the message to the opposite terminal (steps 308 and 309).

  A series of processes starting from acquisition of the reception signal (step 302) is repeatedly executed until the communication path is disconnected.

  It should be noted that, since the pseudo echo creating unit 203 starts the filter coefficient update algorithm using the given delay time as an initial value, whether or not a sudden change in the delay time has occurred based on the given delay time. Judging. When it is determined that a sudden change has occurred in the delay time, the filter coefficient update is reset, and the filter coefficient update algorithm is restarted with the delay time after the determination as an initial value.

(A-3) Effect of the Embodiment According to the above embodiment, the delay time in the echo path is detected based on the signal obtained by filtering the received signal and the transmitted signal into a frequency band having a large individual difference. Therefore, even when the fluctuation range of the delay time in the echo path is wide, the delay time related to the echo path can be detected with high accuracy, the learning accuracy of the echo path state can be improved, and the echo removal accuracy can be improved.

  For example, even when not only the far-end speaker but also the near-end speaker is talking when trying to detect the delay time, the delay time related to the echo path can be detected with high accuracy.

(B) Other Embodiments The above embodiment is intended for a case where the fluctuation range of the delay time in an echo path such as an echo canceller according to a soft phone is wide. For analog phones, hands-free phones, etc. The present invention can be applied to echo cancellers for other telephones.

  In the above embodiment, the reception signal filter 200 and the transmission signal filter 201 have a fixed pass band. However, the pass band can be changed according to the position of the operation element (for example, volume). Also good. Here, the band of the reception signal filter 200 and the transmission signal filter 201 may be variable in conjunction with each other, or may be variable individually. Further, it may be possible to instruct the change of the pass band after the call is started.

  Furthermore, in the above-described embodiment, the delay time detection process is shown in one stage. However, depending on the situation, the detection may be performed in multiple stages. For example, in the delay time detection process based on a signal obtained by filtering the received signal and the transmitted signal to 2 kHz or higher (upper limit frequency (8 kHz)), when the highest correlation value does not exceed the threshold, the received signal The delay time may be detected based on a signal obtained by filtering the transmission signal to 3 kHz to 4 kHz with a narrower band.

  Furthermore, in the description of the above embodiment, the echo canceller is centered on the software including the echo canceller program and the CPU that executes the echo canceller. However, the present invention may be realized by a hardware configuration, Of course, the software configuration and the hardware configuration may be mixed and realized.

It is a block diagram which shows the internal structure of the telephone terminal containing the echo canceller of embodiment. It is auxiliary explanatory drawing of the characteristic of the reception signal filter and transmission signal filter of an embodiment. It is a flowchart which shows operation | movement of the echo canceller of embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 ... Telephone terminal 101 ... Network communication part 102 ... Echo canceller part 103 ... Speaker part 104 ... Microphone part 200 ... Reception signal filter 201 ... Transmission signal filter 202 ... Delay calculation part 203 ... Pseudo echo Creation unit, 204... Echo cancellation unit.

Claims (2)

In the echo canceller that removes the echo component included in the transmission signal using the pseudo echo signal created by the pseudo echo signal creation means having the adaptive filter and the filter coefficient update unit,
A reception signal filter that generates a signal limited to a band having a large individual difference in the voice signal from the reception signal input to the echo canceller;
A transmission signal filter for generating a signal limited to a band having a large individual difference in a voice signal from a transmission signal input to the echo canceller;
Search for a time difference in which the correlation between the signal generated by the reception signal filter and the signal generated by the transmission signal filter is high, detect the delay time in the echo path, and detect the delay time given to the filter coefficient update unit And an echo canceller. Computer
A pseudo echo signal creating means having an adaptive filter and a filter coefficient updating unit for creating a pseudo echo signal;
A reception signal filter that generates a signal limited to a band having a large individual difference in the voice signal from the reception signal input to the echo canceller;
A transmission signal filter for generating a signal limited to a band having a large individual difference in a voice signal from a transmission signal input to the echo canceller;
Search for a time difference in which the correlation between the signal generated by the reception signal filter and the signal generated by the transmission signal filter is high, detect the delay time in the echo path, and detect the delay time given to the filter coefficient update unit An echo canceller program that functions as a means.

Images