JP2008160420A - Echo noise canceller and echo noise canceling method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To allow a coefficient value of a half fixed filter to be appropriately set when providing an adaptive filter and the half fixed filter to remove echo. <P>SOLUTION: An echo noise canceller removes echo noise of a sound signal picked up by a microphone and is provided with the adaptive filter and the half fixed filter. The adaptive filter is a filter which detects an echo component of digital sound data and performs echo canceling processing of removing the detected echo component and has a filter coefficient for echo component detection updated successively. The half fixed filter saves the filter coefficient value set by the adaptive filter, and the half fixed filter sets the saved filter coefficient value to perform echo canceling processing in the case that it has been detected continuously a prescribed number of times that the saved filter coefficient value set by the adaptive filter is appropriate. Thus, echo canceling is performed with a use of the half fixed filter whose filter coefficient is set. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an echo noise canceling apparatus and an echo noise canceling method that are suitable for use in a system that performs a full duplex call in a loudspeaker call system such as a hands-free call in a telephone or a video conference system.

  Conventionally, in a voice call system such as a video conference system, sound collected by a microphone of a far-end device is transmitted to the near-end device and emitted from a speaker of the near-end device. The near-end device also includes a microphone, and the sound collected by the microphone is sent to the far-end device and emitted from the speaker of the far-end device.

  In the case of a system configuration in which a full-duplex call is performed in such a loudspeaker call system, the sound emitted from the speaker at each of the far-end device and the near-end device is input to the microphone. When no processing is performed, this sound is sent again to the device on the other side, causing a phenomenon called “echo” in which one's utterance is output from the speaker with a slight delay as if it is a piece. When the level of the echo sound is high, the echo sound is further collected by the microphone, and the system loops to cause a phenomenon called howling.

  An echo canceling device (echo canceller) is known as a device for preventing such echo and howling. As a conventional general echo canceller, an impulse response between a speaker and a microphone is measured using an adaptive filter. Then, the sound from the speaker input to the microphone is removed by generating a pseudo echo signal obtained by convolving the aforementioned impulse response with the reference signal emitted from the speaker, and subtracting this.

Patent Document 1 discloses an example of an echo cancellation device using an adaptive filter.
JP 2000-151474 A

  However, in a so-called double talk state where a speaker on the near-end device side and a speaker on the far-end device side are simultaneously speaking, the voice of the speaker on the far-end device side is added to the sound emitted from the speaker. Is added, and the sound in such a state is input to the microphone. In such a state, the speech of the far-end device side speaker is added to the microphone input signal as a disturbance for the adaptive filter, and if the learning is continued, the coefficient of the adaptive filter becomes a coefficient value far from the correct one. It will get worse. If processing is performed in such a state where the filter coefficient is deteriorated, echoes cannot be removed as a result.

  As a countermeasure against this double talk, for example, apart from the adaptive filter, a semi-fixed filter composed of a FIR (Finite Impulse Response) type filter is prepared, and when the condition of the coefficient of the adaptive filter is improved, Coefficient values set in the adaptive filter are copied to a semi-fixed filter, and echo cancellation is removed by the semi-fixed filter using the copied coefficient value. Usually, a pseudo echo is generated using a coefficient of the semi-fixed filter, and a process of removing the echo is performed.

  However, when the configuration is such that the coefficient value of the adaptive filter is transferred to the semi-fixed filter in this way, the coefficient value of the adaptive filter is changed to a half when it is determined that the coefficient of the adaptive filter is better than the coefficient of the semi-fixed filter. Although it is transferred to the fixed filter, the coefficients that have been determined multiple times are not all correct coefficients, and the appropriate coefficient values are not always copied to the semi-fixed filter. there were. In particular, since the coefficient of the adaptive filter is constantly updated, there is a possibility that the coefficient value at the time of the last judgment among a plurality of judgments may be incorrect. In such a case, the semi-fixed filter There is a problem that an incorrect coefficient value is copied to the signal, resulting in an echo.

  The present invention has been made in view of such a point, and an object of the present invention is to make it possible to appropriately set the coefficient value of a semi-fixed filter when echo cancellation is performed by providing an adaptive filter and a semi-fixed filter. .

The present invention is applied to an apparatus that removes echo noise when transmitting and / or receiving an audio signal picked up by a microphone. An adaptive filter and a semi-fixed filter are provided.
The adaptive filter is a filter in which an echo component of digital audio data obtained by digitizing an audio signal is detected, an echo cancellation process is performed to remove the detected echo component, and filter coefficients for detecting the echo component are sequentially updated. .
The semi-fixed filter saves the filter coefficient value set by the adaptive filter, and when it is detected that the filter coefficient value set by the adaptive filter is appropriate for a predetermined number of times in succession. This is a filter that sets an evacuated filter coefficient value and performs echo cancellation processing.
And the echo noise removal process at the time of transmission and / or reception is performed by the semi-fixed filter configured as described above.

  By doing in this way, when the coefficient of the adaptive filter whose coefficient value is sequentially updated is better than the semi-fixed filter multiple times, the first coefficient of the multiple times is saved in the semi-fixed filter. Those determined to be highly likely to be correct coefficients continuously are set in the semi-fixed filter. Accordingly, it is possible to set an appropriate filter coefficient that reduces the influence of a single error determination.

  According to the present invention, even if there is a single determination of an error in the filter coefficient, the coefficient set in the semi-fixed filter is not affected, and there is a high possibility that an appropriate filter coefficient is set in the semi-fixed filter. Thus, it is possible to perform an appropriate echo cancellation process with few errors, and to perform highly accurate echo cancellation. Accordingly, it is possible to effectively prevent echo and howling when a full-duplex call is made in a loudspeaker call system.

  Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

  FIG. 1 is a diagram showing an example of the overall configuration according to the present embodiment. As shown in FIG. 1, the system configuration is such that a near-end device 10 and a far-end device 20 are connected by a communication line 30. The near-end device 10 and the far-end device 20 in this example are devices for a video conference system, and are configured to perform full duplex calls. Basically, the near-end device 10 and the far-end device 20 have the same configuration. The video transmission configuration is omitted. Each of the devices 10 and 20 is configured as a communication device that performs a so-called loudspeaking call in which sound is amplified and output from the speakers 11 and 21 and picked up by the microphones 12 and 22, and is output from the speakers 11 and 21 of each device. The voice is picked up by the microphones 12 and 22 as well.

  Each of the devices 10 and 20 may be any communication line as long as the communication line 30 can transmit digital data in both directions between the near-end device 10 and the far-end device 20. For example, a communication line provided by a communication provider such as a communication line connected by Ethernet (registered trademark) or a telephone line, a line via the Internet, or the like is applicable. In the following description, the configuration of the near-end device 10 will be described, but the configuration of the far-end device 20 is also the same, and echo cancellation processing described later is performed in each device.

  As shown in FIG. 1, an audio signal picked up and output by the microphone 12 of the near-end device 10 is converted into digital audio data by an analog / digital converter 14, and the converted digital audio data is converted into a signal processing unit 15. Send to. The signal processing unit 15 of this example is configured by a digital signal processor (DSP), and performs processing for converting input and output audio data into desired data. As one of the processes performed by the signal processing unit 15, a noise removal process for canceling echo noise is performed on the signal input from the microphone 12 side. A configuration for noise removal processing will be described later with reference to FIG.

  The audio data processed by the signal processing unit 15 is sent to the audio codec unit 16 to perform codec processing for transmission. Here, the audio data is converted into a code that is standardized by communication of the video conference system. The converted voice data is sent to the communication unit 17 and transmitted to the far end device 20 via the communication line 30.

  The audio data received by the communication unit 17 of the near-end device 10 from the far-end device 20 via the communication line 30 is subjected to decodec processing by the audio codec unit 16 and then sent to the signal processing unit 15. The signal processing unit 15 handles the supplied reception audio data as data necessary for canceling echo noise described later. The audio data output from the signal processing unit 15 is converted into an analog audio signal by the digital / analog converter 13, and the converted audio signal is output from the speaker 11. In FIG. 1, circuits necessary for analog audio signal processing such as an amplifier and a band filter are omitted.

FIG. 2 is a diagram illustrating a configuration example of the signal processing unit 15. In FIG. 2, the configuration for echo noise cancellation processing is mainly shown, and the configuration of other signal processing is omitted.
The digital audio data picked up by the microphone 12 and converted by the analog / digital converter 14 is supplied to the adaptive filter 51 and the semi-fixed filter 53. Each of the adaptive filter 51 and the semi-fixed filter 53 is configured by an FIR (Finite Impulse Response) type filter. Each of the filters 51 and 53 composed of FIR type filters is subjected to processing for removing echo noise components delayed from the input signal by setting the filter coefficient. In this case, the adaptive filter 51 is a filter whose coefficient values are updated at any time. The coefficient value which seems to be optimal is estimated according to the signal state at that time, and is updated as needed with the estimated coefficient value.
The semi-fixed filter 53 is a filter whose filter coefficient is updated at a predetermined time by control based on the comparison in the filter comparator 54. The filter coefficient value set by the adaptive filter 51 is copied and saved to the save filter at a predetermined time. The filter coefficient value saved in the save filter 52 is set in the semi-fixed filter 53. A specific filter coefficient update process will be described later with reference to the flowchart of FIG.

Next, a processing example of audio data in the near-end device 10 will be described with reference to the flowchart of FIG.
First, in the near-end device 10, the encoded voice data from the far-end device 20 is received by the communication unit 17 via the communication line 30 (step S11). The received audio data is decoded by the audio codec block 16. For example, the sampling frequency is 32 KHz, and one sample is decoded into 16-bit straight PCM digital audio data. The decoded audio data is sent as a reference signal to the signal processing unit 15 configured as a DSP and simultaneously sent to the digital / analog converter 13 (step S12). The analog audio signal obtained by the conversion by the digital / analog converter 13 is supplied to the speaker 11 (step S13) and emitted from the speaker 11 (step S14). At the time of sound emission from the speaker 11, amplification processing is performed by an amplifier (not shown), and the sound is output at a sufficient volume for making a loud call.

  In parallel with the reception processing in steps S11 and S12, the digital audio data picked up by the microphone 12 (step S15) and converted by the analog / digital converter 14 is also sent to the signal processing unit 15 configured as a DSP. Are sent as microphone input signals (step S16). The input of the reference signal to the signal processing unit 15 in step S12 and the supply of the microphone input signal to the signal processing unit 15 in step S16 are always given to the signal processing unit 15 one sample at a time. .

  Then, the signal processing unit 15 uses the coefficient set in the semi-fixed filter 53 to estimate the echo component included in the microphone input signal at the time t that is the target of the echo canceling process according to the following [Equation 1]. The value y (t) is calculated, and the value is subtracted (echo canceling) from the value of the microphone input signal and sent to the voice codec unit 16 (step S17).

Here, w (n) is a coefficient vector (p dimension) of the adaptive filter in the n-th echo canceling processing arranged in the order of {w ₀ , w ₁ ,..., W _p−1 }, x (n ) Represents a vector of time domain data of reference signals arranged in the order of {x _t , x _t−1 , x _t−2 ,..., X _{t− (p−1)} }.

  Next, in the adaptive filter 51 (FIG. 2) in the signal processing unit 15, the estimated value y (t) of the echo component included in the microphone input signal at the time t that is the target of the echo canceling process according to the equation [1] Is calculated (step S18). Then, a value obtained by subtracting the estimated y (t) from the value of the microphone input signal is used as an error e (n), and the coefficient of the (n + 1) th turn is calculated using, for example, the update equation of the following [Equation 2]. The vector w (n + 1) is determined. In the formula 2, μ is an update step size.

  The comparator 54 shown in FIG. 2 determines which of the adaptive filter 51 and the semi-fixed filter 53 is superior in canceling echo, and has a counter for the determination. Used processing is performed. That is, first, it is determined whether or not the count value of the counter is 0 (step S19). If the count value is 0 in this determination, the coefficient of the adaptive filter 51 is copied to the save filter 52 (step S20). Thereafter, the comparator 54 determines whether or not the coefficient value of the semi-fixed filter 53 is more suitable for canceling the echo than the coefficient value of the adaptive filter 51 (step S21). If it is determined by this determination that the coefficient value of the adaptive filter 51 is more suitable, the counter in the comparator 54 is reset to 0 (step S23). When it is determined that the coefficient value of the semi-fixed filter 53 is more suitable, 1 is added to the count value of the counter in the comparator 54 (step S22).

  Thereafter, the count value of the counter in the comparator 54 is determined (step S24), and it is determined whether or not the count value is equal to or greater than a preset threshold value. For example, 40 is set as the threshold here. If it is determined that the count value is greater than or equal to the threshold value, the filter coefficient value held by the evacuation filter 52 is copied and set to the semi-fixed filter 53 (step S25). At this time, the count value of the counter is reset to 0 (step S26).

  Up to this point is the noise cancellation processing, and the audio data subjected to the noise cancellation processing in this way is sent from the communication unit 17 of the near-end device 10 to the far-end device 20 via the communication line 30 (step S27). Such a noise cancellation process shown in the flowchart of FIG. 3 is performed by both the near-end device 10 and the far-end device 20, so that a full-duplex call without echo noise can be performed. The coefficient value update process shown in the flowchart of FIG. 3 is continuously executed as long as bidirectional communication is performed, and an optimum coefficient value corresponding to the state at that time is set.

Here, FIG. 4 shows an outline of a state in which the processing shown in the flowchart of FIG. 3 is performed. FIG. 4A shows a processing state according to this embodiment, and FIG. 4B shows a conventional processing state for comparison.
In this example, “40” is set as the threshold value determined in step S24. As shown in FIG. 4A, in the case of the present embodiment, when the adaptive filter 51 starts to determine that an appropriate coefficient is set, the coefficient value is saved in the save filter 52. After that, the coefficient value of the adaptive filter 51 is sequentially updated. Thereafter, when the comparator 54 determines that the coefficient value is the correct coefficient value 40 times in succession, the coefficient value is saved in at least the save filter 52. It is determined that there is a high possibility that the coefficient value 40 times before is a correct filter coefficient that can properly remove the echo noise. Based on this determination, the coefficient value 40 times before saved in the save filter 52 is set in the semi-fixed filter 53. Therefore, the coefficient value set in the semi-fixed filter 53 can be regarded as a coefficient value that has been evaluated to some extent, and it is possible to satisfactorily prevent an incorrect coefficient value from being set.

  On the other hand, in the conventional processing state shown in FIG. 4B, when it is determined that the coefficient value of the adaptive filter is correct for a certain number of times (for example, 40 times) continuously, the adaptive filter is set to the 40th time. This is a process of copying the coefficient value that has been made to the semi-fixed filter. In this conventional process, if the 40th coefficient value is an incorrect coefficient value, the incorrect coefficient value is set as it is in the semi-fixed filter, and then the coefficient value of the semi-fixed filter is updated. Incorrect processing will continue, which is not preferable.

  As described above, the processing of the present embodiment shown in FIG. 4A is compared with the conventional processing example of FIG. Good echo noise removal processing that eliminates the influence of judgment errors is possible.

  In the embodiment described above, “40” is set as the threshold value for determining the count value of the counter in the comparator 54. However, this value “40” is an example, and other values are set as threshold values. You may make it set as. By setting a large value as the value to be determined, the coefficient value when it is determined that it is continuously correct is set, so there is a high possibility that the correct coefficient value will be set. If there is, the possibility that the value is continuously determined to be correct is reduced due to a temporary determination error and the coefficient value of the semi-fixed filter is updated less frequently. It is preferable to set a correct value.

  Further, in the embodiment described so far, the apparatus for the video conference system is configured to cancel the echo noise. However, if the full-duplex call is performed in the voice call system, the telephone that performs the hands-free call. The present invention can also be applied to other various communication devices such as (wired line telephones or wireless telephones). In addition, a computer program for performing the above-described echo noise canceling process is created, and the computer apparatus and its peripheral devices are assembled as a communication device for performing a similar full-duplex call. You may make it function as a communication apparatus which can implement a program and can perform the same noise cancellation process. In this case, a computer device configured as a near-end device and a computer device as a far-end device may be connected so as to be able to transmit data in both directions via the Internet, for example.

It is a block diagram which shows the example of whole structure by one embodiment of this invention. It is a block diagram which shows the structural example of the principal part by one embodiment of this invention. It is a flowchart which shows the process example by one embodiment of this invention. It is explanatory drawing which shows the outline | summary of the processing state by one Embodiment of this invention compared with a prior art example, (a) is a processing example by one Embodiment, (b) is a processing example by a prior art example.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Near-end device, 11 ... Speaker, 12 ... Microphone, 13 ... Digital / analog converter, 14 ... Analog / digital converter, 15 ... Signal processing part, 16 ... Voice codec part, 17 ... Communication part, 20 ... Far End device, 21 ... Speaker, 22 ... Microphone, 30 ... Communication line, 51 ... Adaptive filter, 52 ... Retraction filter, 53 ... Semi-fixed filter, 54 ... Filter comparator

Claims (3)

In an echo noise canceling device for removing echo noise when transmitting and / or receiving an audio signal picked up by a microphone,
Detecting an echo component of digital audio data obtained by digitizing the audio signal, performing echo cancellation processing to remove the detected echo component, and an adaptive filter in which filter coefficients for detecting the echo component are sequentially updated;
The filter coefficient value set by the adaptive filter is saved, and when the filter coefficient value set by the adaptive filter is detected to be appropriate for a predetermined number of times in the saved state, the saving is performed. A semi-fixed filter that performs the echo cancellation processing,
An echo noise canceling apparatus, wherein the semi-fixed filter performs an echo noise removal process when transmitting and / or receiving. The echo noise canceling device according to claim 1,
A filter coefficient temporary storage unit for temporarily storing the saved filter coefficient value;
An echo noise canceling apparatus, wherein when it is detected that the predetermined number of times is appropriate, the filter coefficient value stored in the filter coefficient temporary storage unit is set in the semi-fixed filter. In an echo noise canceling method for removing echo noise when transmitting and / or receiving an audio signal picked up by a microphone,
An appropriate filter coefficient is sequentially applied to the first filter for an adaptive filter that performs echo cancellation processing for detecting an echo component of digital audio data obtained by digitizing the audio signal and removing the detected echo component. A first filter coefficient setting process to be set;
A determination process for determining whether or not the echo cancellation process performed in the filter coefficient setting process is appropriate;
In the determination process, when it is appropriate for a predetermined number of consecutive times, a second filter coefficient value set in the first filter at the first determination of the consecutive number of times is set in the second filter. Perform filter coefficient setting processing,
An echo noise canceling method, wherein the second filter whose filter coefficient is set in the second filter coefficient setting process is used for an echo noise removal process in the transmission and / or reception.

Images