JP6011188B2 - Echo path delay measuring apparatus, method and program - Google Patents

Description

  The present invention relates to an echo path delay measuring apparatus, method, and program, and can be applied to, for example, an apparatus that uses information on propagation delay time (hereinafter simply referred to as delay time) in the echo path of an acoustic echo.

  In recent years, IP telephones using VoIP (Voice over IP) have become widespread. There is one that realizes this as an application on an information processing apparatus such as a personal computer (PC) or a smartphone, and it is generally called a softphone. In the softphone, a general-purpose sound device (including an A / D converter and a D / A converter) built in the information processing apparatus and a driver of the device are used for voice input / output.

  However, such a driver is operated in time division with other software that operates on the information processing apparatus, so that input / output of audio cannot be processed in a specified time depending on the processing of the other software. Sometimes it happens. As a result, the input / output of the audio signal becomes discontinuous due to interruptions and omissions, and it appears to the echo canceller that the characteristics of the echo path (especially the delay time) have changed. The characteristic change is sudden due to the time-division process with other software processes, and it is difficult to eliminate the echo component included in the audio signal from the microphone.

  As a method for dealing with such a change in echo characteristics, Patent Document 1 proposes a method for devising a hardware configuration.

  However, the technology described in Patent Document 1 requires modification of the information processing device or addition of a special circuit, which may hinder the functional characteristics of a general-purpose information processing device. For the user, it was inconvenient.

  In order to solve such a problem, Patent Document 2 proposes a technique that does not involve hardware change.

  The echo canceller described in Patent Document 2 includes a smoothed speech signal calculation means for calculating a smoothed speech signal (speech signal power information) obtained by smoothing a speech signal, and a smoothed speech signal obtained by smoothing the received signal ( Smooth reception signal calculation means for calculating received signal power information) and delay time information formation for obtaining echo path delay time information reflecting delay characteristics of the echo path based on the correlation between the smooth transmission signal and the smooth reception signal Means.

JP 2004-40589 A JP 2007-189543 A

  However, in the technique described in Patent Document 2, there is a case where the correlation between the smooth transmission signal and the smooth reception signal happens to be large due to the background noise component and is calculated as the echo path delay time. In other words, the echo could not be sufficiently erased.

  Therefore, an echo path delay measuring apparatus, method, and program capable of measuring the echo path delay time more stably are desired.

  According to a first aspect of the present invention, there is provided an echo path delay measuring apparatus for measuring a delay time in an echo path. (1) Transmission pitch characteristic acquisition means for obtaining pitch information related to a transmission signal; Received pitch characteristic acquisition means for obtaining pitch information, and (3) information on delay time in the echo path by searching for a time series part of the pitch related to the received signal similar to the time series of pitch related to the transmitted signal And a correlation analysis means for obtaining.

  According to a second aspect of the present invention, in an echo path delay measurement method for measuring a delay time in an echo path, (1) a transmission pitch characteristic acquisition unit obtains pitch information related to a transmission signal, and (2) a reception pitch characteristic. The obtaining means obtains information on the pitch related to the received signal, and (3) the correlation analyzing means searches for the time series part of the pitch related to the received signal similar to the time series of the pitch related to the transmitted signal, The delay time information in the echo path is obtained.

  The third present invention. An echo path delay measurement program for measuring a delay time in an echo path, comprising: (1) transmission pitch characteristic acquisition means for obtaining pitch information related to a transmission signal; and (2) a reception signal Received pitch characteristic acquisition means for obtaining pitch information, and (3) information on delay time in the echo path by searching for a time series part of the pitch related to the received signal similar to the time series of pitch related to the transmitted signal It is made to function as a correlation analysis means which calculates | requires.

  According to the present invention, since the echo path delay time is measured based on the similarity of the variation in the pitch characteristics of the transmission signal and the reception signal, the echo path delay time can be measured more stably than in the past.

It is a block diagram which shows the functional structure of the echo path delay measuring apparatus of 1st Embodiment. It is explanatory drawing which shows the processing image of the transmission pitch characteristic calculation circuit in the echo path delay measuring apparatus of 1st Embodiment. It is explanatory drawing which shows the process image of the correlation analysis circuit in the echo path delay measuring apparatus of 1st Embodiment. It is a block diagram which shows the functional structure of the echo path delay measuring apparatus of 2nd Embodiment. It is a block diagram which shows the functional structure of the echo path delay measuring apparatus of 3rd Embodiment.

(A) First Embodiment Hereinafter, a first embodiment of an echo path delay measuring apparatus, method and program according to the present invention will be described with reference to the drawings. The echo path delay measuring apparatus according to the first embodiment is intended to be incorporated into an echo canceller apparatus.

  In the first embodiment, the pitch characteristics of each of the received signal (far-end signal) and the transmitted signal (near-end signal) are analyzed, and the echo path delay time is measured based on the similarity (correlation) of both pitch characteristics. To do.

(A-1) Configuration of the First Embodiment FIG. 1 is a block diagram showing a functional configuration of the echo path delay measuring apparatus 100 of the first embodiment.

  The echo path delay measuring device of the first embodiment is mounted on an information processing device such as a PC or a smartphone, for example. The echo path delay measuring apparatus of the first embodiment may be constructed as a dedicated board, for example, and realized by writing an echo path delay measuring program to a DSP (digital signal processor). It may be realized by a CPU and software (echo path delay measurement program) executed by the CPU, but can be functionally represented in FIG.

  In FIG. 1, an echo path delay measuring apparatus 100 according to the first embodiment includes a transmission pitch characteristic calculation circuit 101, a reception pitch characteristic calculation circuit 102, and a correlation analysis circuit 103.

  The echo path delay measuring apparatus 100 according to the first embodiment captures a digitized digital transmission signal (hereinafter referred to as a transmission signal) s (n) captured by a near-end microphone (not shown), and is not shown. A digital reception signal (hereinafter referred to as reception signal) r (n) from the far end toward the speaker is input, and an echo path delay is generated from the transmission signal s (n) and reception signal r (n). Time information T is obtained. It is assumed that the delay time characteristics in the echo path, which is the propagation path of sound and sound, until the microphone captures the sound and sound emitted from the speaker are assumed to be constant. For example, a configuration in which a received signal is processed on the speaker side from the echo path delay measuring apparatus 100 and a configuration in which a transmitted signal is processed on the microphone side from the echo path delay measuring apparatus 100 are software processing configurations by time division processing by the CPU. In some cases, such a delay time characteristic change occurs.

  Further, in the echo path delay measuring apparatus 100 of the first embodiment, the transmission signal s (n) is processed in units of frames cut out for each number of samples for a certain time, and the reception signal r (n) is also processed. The processing is performed in units of frames cut out for each number of samples for a certain period of time. In the following description, the number of samples in one frame is expressed as N.

As shown in the equation (1), the transmission pitch characteristic calculation circuit 101 has the transmission signals s (x− (N−1)) to s (x) of the current frame and arbitrary past positions for each frame. Calculate autocorrelation RS (j) with transmission signals s (x−j− (N−1)) to s (x−j) for one frame, and search for j _max that maximizes autocorrelation. The maximum j _max is output to the correlation analysis circuit 103 as the shift amount Sp of the current frame. Note that PS (j) in the equation (1) is a standard value of power considering the power of both the transmission signal and the reception signal calculated according to the equation (2). The sum Σ in the equations (1) and (2) is for n of x− (N−1) to x. The maximum shift amount j _max is searched in a range where j is approximately 2.5 ms to 18.5 ms. The shift amount j _max is a shift that maximizes the autocorrelation, and therefore reflects the pitch characteristics of the transmitted signal at that time.

  FIG. 2 shows transmission signals s (x− (N−1)) to s (x) of the current frame and transmission signals s (x−j− (N−1) for one frame at an arbitrary position in the past. ) To s (x−j) and an explanatory diagram conceptually showing the relationship between the shift amount j.

RS (j) = Σ (s (n) * s (n−j)) / PS (j) (1)
PS (j) = √ (Σ (s (n) * s (n))
* √ (Σ (s (n−j) * s (n−j))) (2)
The reception pitch characteristic calculation circuit 102 performs the same processing as the transmission pitch characteristic calculation circuit 101 on the reception signal r (n), and sends the deviation Rp of the current frame that maximizes the autocorrelation to the correlation analysis circuit 103. Output.

The correlation analysis circuit 103 stores a time series of the deviation amount Sp related to the transmission signal and the deviation amount Rp related to the reception signal given in units of frames, and these time series Sp (ML + 1) to Sp ( M) and Rp (M−k−L + 1) to Rp (M−k), while changing the parameter k within a predetermined range, the similarity error R (k) is calculated according to the equation (3), Search for _kmin that minimizes the similar error, convert _kmin represented by the number of frames to minimize the similar error, and output to the next stage as the echo path delay time Τ It is. In equation (3), m is a parameter representing the frame order (time) on the time series. The time for calculating the similarity error is L (L is an arbitrary integer greater than or equal to 1; when L is 1, the summation process in equation (3) is not required), and the time of the current frame is M. The sum Σ in equation (3) is for m from M−L + 1 to M. The _kmin that minimizes the similarity error is very similar to the change tendency of the pitch characteristic related to the received signal and the change tendency of the pitch characteristic related to the transmitted signal. It can be understood as a value representing the delay time in the echo path when wrapping around.

  FIG. 3 shows the time series Sp (ML + 1) to Sp (M) of the deviation amount Sp related to the transmission signal and the time series Rp (M−K−L + 1) to Rp (M−) of the deviation amount Rp related to the reception signal. k) It is explanatory drawing which shows notionally the relationship between the transmission signal s (x- (N-1))-s (x) of the present flame | frame, and the parameter k.

R (K) = Σ ((Sp (m) −Rp (m−k))
* (Sp (m) -Rp (m-k)) (3)
Here, when the minimum value of the searched similar errors exceeds a threshold set to a certain value, the correlation analysis circuit 103 determines the time series Sp (ML + 1) to Sp (M) and Rp of the two signals. Assuming that (Mk−L + 1) to Rp (M−k) are not similar, the output of the echo path delay time Τ to the next stage is stopped.

  When the circuit at the next stage is the main body of the echo canceller apparatus, for example, as described in Patent Document 2, an echo path delay time Τ is used.

(A-2) Operation of the First Embodiment Next, the operation of the echo path delay measuring apparatus 100 of the first embodiment will be described with reference to FIG. 1 described above.

  The reception signal r (n) from the far end is sounded and output from a speaker (not shown), and is captured by a microphone (not shown) via the echo path with a delay time T inherent to the echo path at that time. It enters the signal s (n) as an echo component. Since the received signal r (n) is mainly an audio signal, it has a pitch characteristic peculiar to the far-end speaker. For example, a sine wave signal, which is a simple waveform, has a constant pitch, but in the case of an audio signal, the pitch is not constant but varies slowly. Since the received signal r (n) enters the transmitted signal s (n) as an echo, the pitch characteristic of a certain period in the received signal r (n) is delayed by the delay time T of the echo path, and the transmitted signal (echo in Appears as pitch characteristics of component s (n).

  In the echo path delay measuring apparatus 100 of the first embodiment, paying attention to the above points, the pitch characteristic fluctuation in the received signal r (n) similar to the pitch characteristic fluctuation in the transmission signal s (n) is searched. Thus, the delay time T of the echo path is obtained.

The transmission signal s (n) is input to the transmission pitch characteristic calculation circuit 101. In the transmission pitch characteristic calculation circuit 101, as shown in the equation (1), the transmission signal s (x− (N−1)) to s (x) of the current frame and an arbitrary position in the past are shown for each frame. The autocorrelation RS (j) with the transmission signals s (x−j− (N−1)) to s (x−j) for one frame is calculated, and the autocorrelation RS (j) is maximized j _max is output to the correlation analysis circuit 103 as the shift amount Sp of the current frame.

The reception signal s (n) is input to the reception pitch characteristic calculation circuit 102. In the reception pitch characteristic calculation circuit 102, for each frame, reception signals r (x− (N−1)) to r (x) of the current frame and reception signals r (x−) of one frame at any past position. j- (N−1)) to r (x−j) are calculated (see equation (1)), and j _{max at} which autocorrelation RR (j) is maximized is the current frame. Is output to the correlation analysis circuit 103 as a deviation amount Rp.

In the correlation analysis circuit 103, the time series of the shift amount Sp related to the transmitted signal and the shift amount Rp related to the received signal given in frame units are stored. Then, in the correlation analysis circuit 103, the parameter k is set to a predetermined range between these time series Sp (ML + 1) to Sp (M) and Rp (MkL + 1) to Rp (Mk). , The similar error R (k) is calculated according to the equation (3), and the _kmin represented by the number of frames that minimizes the similar error is converted into milliseconds (ms), and the echo path delay time is calculated. It is output to the next stage as Τ.

(A-3) Effect of the First Embodiment According to the first embodiment, the delay time of the echo path is measured based on the time variation of the pitch representing the characteristics of the voice signal (received signal). Compared to the case of measuring based on the power fluctuation described in Patent Document 2, it is less affected by background noise and the like, and a stable echo path delay time measurement result can be obtained.

(B) Second Embodiment Next, a second embodiment of the echo path delay measuring apparatus, method and program according to the present invention will be described with reference to the drawings.

  FIG. 4 is a block diagram showing a functional configuration of the echo path delay measuring apparatus 100A of the second embodiment.

  In FIG. 4, an echo path delay measuring apparatus 100 </ b> A according to the second embodiment includes a transmission side LPC (Linear Predictive Coding) analysis in addition to a transmission pitch characteristic calculation circuit 101, a reception pitch characteristic calculation circuit 102, and a correlation analysis circuit 103. A circuit 104 and a receiver LPC analysis circuit 105;

  The transmission signal s (n) is input to the transmission side LPC analysis circuit 104. The transmission-side LPC analysis circuit 104 performs LPC analysis on the transmission signal s (n), and among the LPC coefficients and the residual signals obtained by the analysis, the residual signal is sent to the transmission pitch characteristic calculation circuit 101. Give.

  On the other hand, the reception signal r (n) is input to the reception side LPC analysis circuit 105. The reception side LPC analysis circuit 105 performs LPC analysis on the reception signal r (n), and gives a residual signal to the reception pitch characteristic calculation circuit 102 among the LPC coefficient and the residual signal obtained by the analysis. is there.

  Unlike the first embodiment, the transmission pitch characteristic calculation circuit 101 does not use the transmission signal s (n) but the residual signal from the transmission side LPC analysis circuit 104 in the first embodiment. The processing described is performed.

  Similarly, the reception pitch characteristic calculation circuit 102 differs from the first embodiment with respect to the residual signal from the reception side LPC analysis circuit 105 instead of the reception signal r (n) in the first embodiment. The processing described is performed.

  As is well known, the residual signal is generally an excitation source signal (vocal cord drive signal) obtained by excluding vocal tract information from a voice signal (transmitted signal or received signal), and the pitch becomes clearer than the voice signal itself. It is what. Therefore, in the echo path delay measuring apparatus 100A of the second embodiment, the pitch characteristic is detected from the residual signal with respect to the transmission signal and the reception signal.

  Except for the above points, the second embodiment is the same as the first embodiment.

  According to the second embodiment, the pitch characteristics of the transmission signal and the reception signal can be detected more accurately than in the first embodiment, and as a result, the echo path is less affected by background noise and the like, and is more stable. A measurement result of the delay time can be obtained.

(C) Third Embodiment Next, a third embodiment of the echo path delay measuring apparatus, method and program according to the present invention will be described with reference to the drawings.

  FIG. 5 is a block diagram showing a functional configuration of the echo path delay measuring apparatus 100B of the third embodiment.

  In FIG. 5, an echo path delay measuring apparatus 100 </ b> B according to the third embodiment includes a sound / silence determination circuit 106 in addition to a transmission pitch characteristic calculation circuit 101, a reception pitch characteristic calculation circuit 102, and a correlation analysis circuit 103.

  The voice signal s (n) is given to the voice / silence determination circuit 106. The sound / silence determination circuit 106 performs sound / silence determination for each frame and gives the determination result to the correlation analysis circuit 103. As the sound / silence determination method applied by the sound / silence determination circuit 106, any existing method may be applied. Here, as a threshold for separating voice or silence, a threshold for determining “sound” even when the near-end speaker's voice signal includes only an echo signal that is not included in the transmission signal. It is preferable to apply.

  The correlation analysis circuit 103 according to the third embodiment stores a time series of the sound / silence determination result for each frame given from the sound / silence determination circuit 106. When the current frame is silent, the correlation analysis circuit 103 does not perform analysis and does not output the echo path delay time T (note that the echo path delay time T immediately before effective analysis is output). Is also good). When the current frame is sound, the correlation analysis circuit 103 grasps a section composed of one or a plurality of frames in which the sound continues from the current frame to the past, and the transmission system of the sound continuous frame section Is recognized by the method described in the first embodiment, and an echo path delay time T is obtained.

  Except for the above points, the third embodiment is the same as the first embodiment.

  According to the third embodiment, since the echo path delay time is measured on condition that at least the transmitted signal is sound, as a result, the echo path delay time is hardly affected by background noise and the like, and is more stable. The measurement result of the echo path delay time can be obtained.

(D) Other Embodiments In the description of each of the above embodiments, various modified embodiments have been referred to, but further modified embodiments as exemplified below can be given.

  You may make it use together the technical idea which added the LPC analysis circuit in 2nd Embodiment, and the technical idea which added the sound / silence determination circuit in 3rd Embodiment.

  In the second embodiment, the LPC analysis is used to obtain the excitation source signal (voice band driving signal). However, the excitation source information may be obtained by other methods.

  In the second embodiment, the transmission signal processing system and the reception signal processing system are provided with the LPC analysis circuit, but only one (for example, the transmission signal processing system) performs the LPC analysis. A circuit may be provided.

  In the third embodiment, the voice / silence determination is performed only for the transmission signal. However, the voice / silence determination may be performed for the reception signal. For example, you may make it obtain | require the gap frame number with which a similar error becomes the minimum between the pitch characteristic fluctuation | variation in the sound area of a transmission signal, and the pitch characteristic fluctuation | variation in the sound area of a reception signal.

  The method for detecting the pitch of the transmitted signal and the received signal is not limited to those of the above embodiments, and an index indicating the similarity between the pitch characteristic variation of the transmitted signal and the pitch characteristic variation of the received signal is also provided. Of course, it is not limited to the similar error.

  In each of the above embodiments, an echo canceller apparatus related to a softphone on a PC or a smartphone is intended, but the application of the present invention is not limited thereto. Not only can it be applied as an echo canceller for devices where the characteristics (especially delay characteristics) of the echo path may change suddenly, but it can also be applied as an echo canceller for devices where the change in the characteristics of the echo path is gradual. be able to. The target echo is not limited to an acoustic echo. For example, in a PC, an audio signal that flows through a signal line may cause crosstalk that flows to other signal lines. Such crosstalk is also a kind of echo, and the delay described in the above embodiments. Time measurement methods can be applied. The claims also include echoes that are considered such echoes.

  DESCRIPTION OF SYMBOLS 100, 100A, 100B ... Echo path delay measuring apparatus, 101 ... Transmission pitch characteristic calculation circuit, 102 ... Reception pitch characteristic calculation circuit, 103 ... Correlation analysis circuit, 104 ... Transmission side LPC analysis circuit, 105 ... Reception side LPC analysis Circuit 106: Sound / silence determination circuit.

Claims (5)

In an echo path delay measuring device for measuring a delay time in an echo path,
Transmission pitch characteristic acquisition means for obtaining pitch information related to the transmission signal;
A reception pitch characteristic acquisition means for obtaining pitch information related to the reception signal;
An echo path comprising: correlation analysis means for obtaining delay time information in the echo path by searching for a time series part of the pitch related to the received signal similar to the time series of the pitch related to the transmitted signal Delay measurement device. The transmission pitch characteristic acquisition means is
A first excitation source signal obtaining unit for obtaining a first excitation source signal obtained by removing vocal tract information from a transmission signal;
A transmission pitch characteristic acquisition unit that obtains pitch information in the first excitation source signal as pitch information related to the transmission signal;
The reception pitch characteristic acquisition means includes:
A second excitation source signal acquisition unit for obtaining a second excitation source signal obtained by removing vocal tract information from the received signal;
The echo path delay measuring apparatus according to claim 1, further comprising: a received pitch characteristic acquisition unit that obtains pitch information in the second excitation source signal as pitch information related to the received signal. A voice / silence determination means for performing voice / silence determination on the transmission signal;
The echo path delay measuring apparatus according to claim 1 or 2, wherein the correlation analysis unit obtains information on a delay time in the echo path on condition that the transmitted signal is voiced. In the echo path delay measurement method for measuring the delay time in the echo path,
The transmission pitch characteristic acquisition means obtains pitch information related to the transmission signal,
The received pitch characteristic acquisition means obtains information on the pitch related to the received signal,
An echo path delay characterized in that the correlation analysis means obtains delay time information in the echo path by searching for a time series part of the pitch related to the received signal similar to the time series of the pitch related to the transmitted signal Measuring method. An echo path delay measurement program for measuring a delay time in an echo path,
Computer
Transmission pitch characteristic acquisition means for obtaining pitch information related to the transmission signal;
A reception pitch characteristic acquisition means for obtaining pitch information related to the reception signal;
It is made to function as a correlation analysis means for obtaining delay time information in the echo path by searching for a time series part of the pitch related to the received signal similar to the time series of the pitch related to the transmitted signal. Echo path delay measurement program.

Images