JP2010118793A - Propagation delay time estimator, program and method, and echo canceler - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce the frequency of defining error time as an estimation result in the estimation of propagation delay time. <P>SOLUTION: A propagation delay time estimator estimates propagation delay time that is a time difference between a first discrete time signal and a second discrete time signal which are originally the same signal. The propagation delay time estimator includes: a means for converting the first discrete time signal into a third discrete time signal which is expressed by the smaller number of quantization bits and converting the second discrete time signal into a fourth discrete time signal which is expressed by the smaller number of quantization bits; and a means for calculating an estimation value of the propagation delay time by using the third and fourth discrete time signals. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a propagation delay time estimator, method and program, and an echo canceller, and can be applied to, for example, an echo canceller that cancels a line echo.

  FIG. 8 shows an arrangement example of a line echo canceller having a large round-trip propagation delay time (hereinafter also simply referred to as “propagation delay time”) to the echo generation point. In FIG. 8, the two-wire transmission lines 401, 402, 403, and 404 are drawn with one line for convenience.

  A transmission path 401 is a propagation path of an audio signal x (n) transmitted from a telephone on the opposite side (not shown) facing the telephone 408, and is converted into a two-line / four-line via the line echo canceller 405 and the relay network 406. Connected to the device 407. The audio signal x (n) reaches the telephone 408 via the 2-wire / 4-wire converter 407, and at the same time, part of it due to impedance mismatch at the 2-wire / 4-wire converter 407. Is reflected to the line echo canceller 405 via the two-line transmission path 403 and the relay network 406 as a signal y (n). The echo signal y (n) is erased by the line echo canceller 405. However, since the echo signal y (n) cannot generally be completely erased, the residual echo signal e (n) is transmitted to the opposite side via the two-line transmission line 404. Go back to the phone.

  Next, a learning identification method that is a typical echo cancellation algorithm generally used in the line echo canceller 405 will be described. Hereinafter, unless otherwise specified, the signal is regarded as a discrete value signal quantized at a certain sampling rate, and the time is expressed as a discrete time n with a sampling period as one unit time.

In the echo cancellation algorithm, the echo path on the telephone 408 side viewed from the line echo canceller 405 is regarded as a linear shift invariant system, specifically, an Nth order finite impulse response filter. The tap coefficient of the filter is h (k), the tap coefficient at time n of the filter estimated by the line echo canceller 405 is g _n (k), and the echo signal estimated by the line echo canceller 405 is y ′. When expressed as (n), the algorithm can be expressed by equation (1).

In the equation (1), μ is a parameter that determines the convergence speed of g _n (k), and is usually set in a range of 0 <μ <2.

Next, the round-trip propagation distance between the line echo canceller 405 and the 2-line / 4-line converter 407, which is the echo generation point, is long, and due to processing delays by various network devices arranged inside the relay network 406. The round-trip propagation delay time T _rt between the line echo canceller 405 and the 2-wire / 4-wire converter 407 is assumed to be sufficiently large. Here, T _rt corresponds to _NDLY unit time in terms of discrete time, and when this delay is expressed by a transfer function _HDLY (z) based on the z function, it is expressed by equation (2).

Further, when the 2-wire / 4-wire converter 407 is regarded as a finite impulse response filter, the filter order is N _HYB , and the tap coefficients of the filter are h (0), h (1),..., H (N _HYB ), And the transfer function based on the z function of the filter is represented by H _HYB (z), the result is as shown in equation (3).

Therefore, when the transfer function by the z function of the echo path at this time is expressed as H _EP (z), it is as shown in the equation (4). If the input / output signals of the transfer function of the echo path expressed by the equation (4) are x (n) and y (n), respectively, and these z-transforms are X (z) and Y (z), then (5) The relationship shown in the equation is established. In equation (5), z- ^NDLY X (z) means that the input signal x (n) is delayed by _NDLY unit time in the time domain. Actually, when the equation (5) is converted into the time domain by inverse z transformation, the equation (6) is obtained.

FIG. 9 is an explanatory diagram showing a specific example of a signal h _EP (n) in the discrete time domain obtained by inversely z-converting H _EP (z).

The temporal behavior of h _EP (n) follows the above description and is expressed by the equation (7).

Here, the filter order N of the estimator of the line echo canceller 405 that performs echo estimation according to the equation (1) is assumed, and the relationship between N _DLY and N _HYB is assumed as the equation (8), and the equation (6) The echo signal y (n) represented and the echo estimation signal y ′ (n) represented by the equation (1) are compared as shown in the equation (9).

In equation (9), the coefficient g (k) of x (n) on the right side of the estimated echo signal y ′ (n) is the coefficient h (k) of x (n) on the right side of the echo signal y (n) by learning. Converge as shown in equation (10).

Equation (10) suggests the following. If the propagation delay time of the echo path is known in advance, the line echo canceller 405 can omit the calculation of the filter coefficients g (0) to g (N _DLY −1) and at the same time g (0) to g A memory area for storing (N _DLY −1) can also be saved. Actually, the impulse response time of the 2-wire / 4-wire converter 407 is found to be within 16 msec according to the actual measurement result by Non-Patent Document 1, whereas the propagation delay time of the echo path is the network topology. In addition, it depends on the equipment operation policy of the telecommunications carrier, and it is impossible to know in advance. Therefore, in this case, it is desirable that the line echo canceller has a function capable of estimating the delay of the echo path. This makes it possible to reduce the amount of processing related to the echo estimation of the line echo canceller, and is a great advantage especially for a device that accommodates a large number of voice calls such as a VoIP (Voice over IP) gateway device for a communication carrier.

[Propagation delay time estimation by conventional technology]
Conventional methods for estimating the delay time of the echo path include the following.

  The first conventional method is disclosed in Patent Document 1 or the like, and this method adds time information to a voice packet, thereby calculating a propagation delay time. Regarding this method, in IP (Internet Protocol), a mechanism for calculating the round-trip propagation delay time in a more general manner is defined by Non-Patent Document 2. As a result, the line echo canceller need not include a propagation delay time estimator.

  The second conventional method is disclosed in Patent Document 2, Patent Document 3, Patent Document 4, and the like, and performs normal echo estimation processing without providing a function dedicated to propagation delay time estimation. That is, the echo path is regarded as a linear shift invariant system, and the transfer function of the echo path is estimated using an algorithm such as the learning identification method described above. Specifically, the transfer function estimation coefficient g ( k) is calculated, and if, for example, converges to g (k) ≈0 in the range of 0 ≦ k ≦ d, this d is regarded as the propagation delay time of the echo path. An estimation coefficient regarded as a pure delay is excluded from the echo estimation process, thereby reducing the processing amount of the echo canceller.

The third conventional method is disclosed in Patent Document 5 and Patent Document 6, and obtains a cross-correlation between an audio signal transmitted by the echo canceller to the telephone side and an echo signal reflected and returned from the telephone side. It determines the propagation delay time.
JP 2001-333000 A JP-A-9-55687 JP-T-2001-501413 Japanese Patent Application Laid-Open No. 7-282859 Japanese Patent Laid-Open No. 7-303061 JP 2005-528039 Gazette ITU-T Recommendation G. 168 Appendix II IETF RFC 3550, RTP / RTCP

  The application condition of the stipulating method of Non-Patent Document 2 relating to the first conventional method is that the propagation delay time of the part where the echo is generated and the part where the voice signal is packetized can be ignored, and This is limited to the case where the interface is an IP packet interface. However, it is extremely rare that an IP network is interposed between the echo canceller and the echo reflection point. In addition, this method cannot be applied to a line echo canceller mounted on a VoIP apparatus that generally accommodates existing fixed telephones with a time division multiplexing (TDM) interface.

  In addition, the second conventional method cannot reduce the amount of processing in the initial echo estimation learning at all, and this cannot withstand the specification of the VoIP gateway device that accommodates a large number of telephone calls. The reason is that in these communication apparatuses, the time required to change the accommodated call to the call state simultaneously for all channels is regarded as an important factor for determining the service operating rate. Therefore, when the second conventional method is adopted in these communication apparatuses, the processing amount is sufficient to perform echo estimation including the round-trip propagation delay time of the echo path simultaneously for all channels, or the processing amount is reduced accordingly. When it cannot be secured, the accommodated call is divided into several groups and divided into several times in time series to perform echo estimation. In any case, the second conventional method is not suitable for an echo canceller mounted on a device that accommodates a large number of voice calls.

  In the third conventional method, it is necessary to obtain a cross-correlation between an audio signal transmitted from the echo canceller to the telephone side and an echo signal reflected and returned from the telephone side. In general, the processing amount of cross-correlation calculation is not so different from that of echo estimation processing, and Patent Document 5 and Patent Document 6 do not mention a specific processing method of correlation calculation.

  The present invention has been made in view of the above points, and an object of the present invention is to efficiently estimate the propagation delay time with a small amount of processing and small hardware.

  The propagation delay time estimator according to the first aspect of the present invention (1) estimates a propagation delay time that is a time difference between a first discrete time signal and a second discrete time signal that were originally the same signal. In the propagation delay time estimator, (2) converting the first discrete time signal into a third discrete time signal expressed by a smaller number of quantization bits, and converting the second discrete time signal to Quantization bit number conversion means for converting to a fourth discrete time signal expressed by a smaller number of quantization bits; (3) the third discrete time signal and the fourth discrete time signal; And a delay time calculating means for calculating an estimated value of the propagation delay time.

  The propagation delay time estimation program of the second aspect of the present invention (1) estimates a propagation delay time which is a time difference between a first discrete time signal and a second discrete time signal which were originally the same signal. In the propagation delay time estimator, (2) converting the first discrete time signal into a third discrete time signal expressed by a smaller number of quantization bits, and converting the second discrete time signal to Quantization bit number conversion means for converting to a fourth discrete time signal expressed by a smaller number of quantization bits; (3) the third discrete time signal and the fourth discrete time signal; It is used to function as delay time calculation means for calculating the estimated value of the propagation delay time.

  The propagation delay time estimation method of the third aspect of the present invention is (1) estimating a propagation delay time that is a time difference between a first discrete time signal and a second discrete time signal that were originally the same signal. In the propagation delay time estimation method, (2) a quantization bit number conversion unit and a delay time calculation unit are provided. (3) The quantization bit number conversion unit converts the first discrete time signal into a smaller number of quantum signals. Converting to a third discrete time signal expressed by the number of quantized bits, converting the second discrete time signal to a fourth discrete time signal expressed by a smaller number of quantization bits, ( 4) The delay time calculating means includes calculating an estimated value of the propagation delay time using the third discrete time signal and the fourth discrete time signal.

  According to the present invention, in the estimation of the propagation delay time, it is possible to reduce the frequency at which an erroneous time is used as an estimation result.

(A) Main Embodiment Hereinafter, an embodiment of a propagation delay time estimator, a program, and a method according to the present invention will be described in detail with reference to the drawings.

(A-1) Configuration of Embodiment In the following description, it is assumed that the propagation delay time estimator of this embodiment is provided in a telephone communication system in association with a line echo canceller.

  FIG. 1 is a block diagram showing the detailed configuration of the propagation delay time estimator of this embodiment and the positioning in the telephone communication system. In FIG. 1, the telephone communication system 1 according to the embodiment includes a propagation delay time estimator 10, an echo canceller 11, and a pseudo delay device 12.

  In the following description, although not shown in FIG. 1, the speaker of the telephone terminal connected to the echo path 13 is referred to as “near-end speaker”. On the other hand, similarly, although not shown in FIG. 1, an opponent of the near-end speaker is referred to as a “far-end speaker”. A signal transmitted to the echo path 13 via the two-wire transmission path 14 is also referred to as a “far end signal”. A signal that flows through the two-wire transmission line 15 on the near end side and is sent to the opposite speaker is referred to as a “near end signal”.

  The line echo canceller 11 sends the far-end signal sent to the echo path 13 via the two-line transmission path 14, reflected by the echo path 13 and returned via the two-line transmission path 15 from the near-end signal. To be removed. The line echo canceller 11 generates a pseudo echo using an adaptive filter or the like based on a far-end signal given through a pseudo delay device 12 described later, and removes the echo using the generated pseudo echo.

  The propagation delay time estimator 10 sends a time difference when the far-end signal is transmitted to the echo path 13 via the two-wire transmission line 14 and the echo is reflected and returned via the two-wire transmission line 15, that is, The propagation delay time of the echo path 13 is estimated. Then, the propagation delay time estimator 10 notifies information related to the estimated propagation delay time to a pseudo delay device 12 described later.

  The propagation delay time estimator 10 includes decoders 111 and 112, absolute value converters 121 and 122, low-pass filters 131 and 132, decipherers 141 and 142, level discriminators 151 and 152, and a temporary storage unit 161. 162, a least square operator 170, a significance determiner 180, and a determination protector 190. The propagation delay time estimator 10 may be a program execution configuration such as a CPU, ROM, RAM, EEPROM, hard disk or the like (not limited to one, but may be configured so that a plurality of units can be distributed. ) May be constructed by installing the propagation delay time estimation program or the like of the embodiment, and even in that case, it can be functionally shown in FIG. Further, the propagation delay time estimator 10 may realize all the components by hardware, or may realize some components by using a program (software) as described above.

  Next, the decoders 111 and 112 will be described.

  The decoder 112 converts the far-end signal given via the two-wire transmission line 14 into a linear code applicable to addition / subtraction / division / division operations that are frequently performed in digital signal processing, This is given to the absolute value converter 122. The decoder 111 converts the near-end signal given via the two-wire transmission line 15 into a linear code, like the decoder 112, and gives the converted signal to the absolute value converter 121.

  In the existing telephone network, ITU-T Recommendation G. In general, transmission is performed using a PCM (Pulse Code Modulation) code complying with H.711. The decoders 111 and 112 convert the encoded speech signal into a linear code that can be added, subtracted, multiplied, and divided, for example, a two's complement fixed point. As a decoding process in the decoders 111 and 112, for example, ITU-T Recommendation G. Since the PCM code of 711 is about 8-bit code, a decoding table can be obtained by preparing a conversion table of 256 words, that is, a conversion ROM, using the PCM code as an address, and storing a linear code at the address. This is completed by simply reading the conversion ROM. Although it may be obtained by calculation according to a decoding algorithm, it is considered that it is cheaper to prepare a conversion ROM in view of today's hardware integration technology and cost per gate.

  Also, the decoders 111 and 112 are ITU-T recommendation G.264. In the case of encoding other than 711, the decoding process may be performed in accordance with an algorithm defined by the encoding.

  Next, the absolute value converters 121 and 122 will be described.

  The absolute value converters 121 and 122 convert the signals given from the decoders 111 and 112 into absolute values, that is, only when the given signals are negative numbers, the magnitude values are equal to each other and the sign is inverted. When the signal is converted and the given signal is a positive number, no particular conversion is performed. The absolute value converters 121 and 122 give the absolute value signals to the low-pass filters 131 and 132.

  In this embodiment, the propagation delay time is calculated by the least squares calculator 170 described later for the envelope signal of the far end signal and the envelope signal of the near end signal. As a simple calculation method of the envelope of the audio signal, it can be obtained by extracting the local maximum value of the audio signal and connecting them. Since this is realized by taking the absolute value of the audio signal and passing it through the low-pass filter, the propagation delay time estimator 10 uses the absolute value converters 121 and 122 to extract the envelope. And low-pass filters 131 and 132 to be described later.

  Next, the low-pass filters 131 and 132 will be described.

  The low-pass filters 131 and 132 block the high-frequency component of the signal from the absolute value signals given from the absolute value converters 121 and 122, and supply them to the thinning-out devices 141 and 142.

  The low-pass filters 131 and 132 allow signals lower than a preset cutoff frequency to pass, and blocks signals higher than that frequency. It is provided for extracting the envelope signal together with the absolute value converters 121 and 122 described above.

  Next, the thinning devices 141 and 142 will be described.

  The decimation devices 141 and 142 decimate part of the signals given from the low-pass filters 131 and 132 to generate a signal with a reduced sampling frequency (sampling speed), and the level discriminator 151. , 152.

  Since the high frequency component is removed from the envelope signal as compared with the original signal, the sampling frequency is effectively lowered, and the decimation devices 141 and 142 decrease the sampling frequency by 1 every several samples. One specimen can be thinned out. In general, the band of the voice envelope signal is about 100 Hz, and the effective sampling frequency is reduced. Therefore, it is possible to thin out a plurality of envelope signals into one envelope signal. For example, assuming that the sampling frequency of a telephone voice signal is generally about 8 kHz, and the maximum effective sampling frequency of an envelope signal is 200 Hz, 40 envelope signals can be thinned out into one envelope signal. It becomes possible. In the following description, the sample period (sample time) after being thinned by the thinning devices 141 and 142 is represented as “M”.

  Next, the level discriminators 151 and 152 will be described.

The level discriminators 151 and 152 discriminate when the envelope signal is larger than the set threshold value as positive and when it is smaller as negative. FIG. 2 shows an example in which the threshold value is set to a value close to zero and is determined to be negative when the state is close to silence.

  FIG. 2 is an explanatory diagram showing an example in which the level discriminators 151 and 152 simply perform level determination based on a threshold value.

  As shown in FIG. 2, the level discriminators 151 and 152 may determine whether the audio signal is positive or negative with reference to a threshold (hereinafter referred to as “TH”). For example, the audio signal may be positive when it is greater than TH and negative when less than TH. Here, for example, when the level changes from a negative value to a positive value, or from a positive value to a negative value, the threshold value to be referred to may be changed so that the threshold value has hysteresis characteristics. . Note that the value of the threshold TH applied between the far-end signal and the near-end signal may be different.

  FIG. 3 is an explanatory diagram showing an example of performing level determination using hysteresis characteristics in the level discriminators 151 and 152.

  For example, as shown in FIG. 3, a sound signal whose signal level is less than −TH transitions to + TH or more and transitions to −TH or less is positive, or a sound signal whose signal level exceeds + TH is − The time between the transition to TH or less and the transition to + TH or more may be negative.

  Next, the temporary storage units 161 and 162 will be described.

Temporary storage units 161 and 162 store information related to signals given from level discriminators 151 and 152 for a certain period of time, and supply the stored information (signals) to least squares calculator 170, respectively.

  FIG. 4 is an explanatory diagram showing a configuration example of the temporary storage unit 161 on the near end side.

  As shown in FIG. 4, the temporary storage unit 161 holds the output of the level discriminator 151 for every M samples, that is, holds the sample value of the LM sample time. You may build using. The M sample time corresponds to the time resolution of the signal in the propagation delay time estimator 10. In FIG. 4, S (n) represents the sample value of the near-end signal from time n to N sample times before.

  FIG. 5 is an explanatory diagram showing the configuration of the temporary storage unit 161 in FIG. 4 and the time relationship of the level determination result by the level determiner 151 given to the temporary storage unit 161.

  FIG. 6 is an explanatory diagram showing a configuration example of the temporary storage unit 162 on the far end side.

  The temporary storage unit 162 is a shift register that holds (L + K−1) outputs of the level discriminator 152 for each M samples, that is, a (L + K−1) M sample time, or a circular queue. The structure of this memory element is shown in FIG. In FIG. 6, R (n) represents a sample value of the far-end signal from time n to N sample times before.

  Next, the least squares calculator 170 will be described.

  The least squares calculator 170 calculates a time difference (propagation) between the near-end signal envelope signal stored in the temporary storage unit 161 for a certain period of time and the far-end signal envelope signal stored in the temporary storage unit 162 for a certain period of time. (Estimated delay time) is calculated, and the calculated result is given to the significance determiner 180. For example, the least squares calculator 170 calculates the square difference of the level discrimination results by the level discriminators 151 and 152 using the time difference as a variable. In this case, the time difference when the square difference between the level discrimination results of the two becomes the minimum (hereinafter referred to as “the least square difference”) corresponds to the propagation delay time of the echo path as viewed from the delay time estimator.

  Hereinafter, a specific example of a method for calculating the estimated value of the time difference (propagation delay time) in the least squares calculator 170 will be described. In the following description, time is expressed as discrete time n, and sample time M is defined as 1 unit time.

When the square difference D ² (n, τ) between S (n) and R (n) at time n is calculated using the time difference τ between S (n) and R (n) as a parameter, the following equation (11) is obtained. It becomes like this.

As shown in the following equation (12), τ (hereinafter referred to as “τ _min ”) that minimizes “D ² (n, τ), 0 ≦ τ ≦ K−1” at time n is a time difference (propagation). Delay time). As described above, the least squares calculator 170 calculates a value related to the difference between S (n) and R (n) for each time difference (τ), for example, as in the following Expression (11), and the calculation result The time difference (propagation delay time) may be estimated based on

^{"D 2 (n, τ),} 0 ≦ τ ≦ K-1 " each time the n is updated, but will calculate the, here, it is assumed that the calculated efficiently D 2 ^(n, τ) . D ² (n, τ) at time n + 1 can be expressed as the following equation (13), similarly to the above equation (11).

In S (n) and R (n), the result of the level discriminators 151 and 152, that is, the determination result of positive or negative, is held. The positive determination result is 1, and the negative determination result is 0. In this case, it is possible to transform the above (11) into the following equation (14) and the above (13) into the following equation (15).

However, there are many common terms in the sum Σ of the equations (14) and (15). If the above equation (15) is re-expressed while focusing on only different terms, the following equation (16) is obtained. This makes it possible to realize the N-square difference sum calculation only by addition. The least squares calculator 170 may use the following equation (16) in calculating τ _min .

  Next, the significance determiner 180 will be described.

The significance determiner 180 determines the significance of the calculation when the result of the calculation is given from the least square calculator 170. The following 1st-4th factors are assumed as a factor which reduces the significance of a calculation. When the significance determination unit 180 passes all the significance tests related to the following first to fourth factors, the calculation result by the least squares calculation unit 170 is considered significant, and the propagation of the echo path 13 at this time is propagated. Τ _min corresponding to the delay time is given to the judgment protector 190. The significance determiner 180 may perform other tests in addition to the significance tests related to the following first to fourth factors. In the first embodiment, the significance determiner 180 regards the calculation result by the least squares calculator 170 as significant when all of the significance tests related to the first to fourth factors are passed. However, the calculation result may be considered significant when passing a part of the significance test relating to the following first to fourth factors.

  The first factor determined by the significance determiner 180 is the far-end speaker's voice envelope signal power (far-end signal envelope signal power). The power of the far-end speaker's voice envelope signal input to the propagation delay time estimator 10 via the two-wire transmission line 14 of FIG. 1, that is, the far-end signal envelope that is the output of the low-pass filter 132. When the signal value is small, the signal-to-noise ratio (S / N ratio) of the far-end speaker envelope signal is deteriorated and the reliability of the operation result of the least squares operator 170 is deteriorated. Need to be invalidated. In FIG. 1, it is assumed that the output of the low-pass filter 132 is monitored by providing a threshold value.

  The second factor determined by the significance determiner 180 is a double talk state.

  The double talk state means that the far-end speaker and the near-end speaker are in a busy state at the same time. As a matter of course, in this case, the calculation for calculating the delay time of the echo path 13 does not make sense, so the calculation result of the least squares calculator 170 needs to be invalidated. For example, the detection of the double talk state is based on the ratio of the output (DC (S)) of the low-pass filter 131 on the near end side to the output (DC (R)) of the low-pass filter 132 on the far end side. An appropriate threshold value may be provided for determination. For example, ITU-T Recommendation G. In 168, since it is assumed that the echo return loss amount of the echo path is 6 dBm or more, assuming that this relationship can be applied to the envelope signal of the voice, in this case, “DC (S)> (1/2 ) · DC (R) ”is established, it can be regarded as a double talk state. Note that the value actually employed as the interval value may be set to a value corresponding to the echo reflection attenuation amount specified by each communication carrier in the subscriber circuit.

  The third factor determined by the significance determiner 180 is the minimum square difference calculated by the least square calculator 170.

Every time the time n is updated, the least squares calculator 170 calculates the above equation (16) and the like for 0 ≦ k ≦ M and gives M D (n, τ) to the significance determiner 180. . Of these, τ _min which is the value of τ that minimizes D (n, τ) corresponds to the delay time to be obtained. However, when D (n, τ) exceeds the threshold value, the correlation between the far-end signal and the near-end signal is regarded as insufficient, and the calculation result of the least squares calculator 170 may be discarded.

  The fourth factor determined by the significance determiner 180 is the periodicity of the envelope signal. In the level discriminator 152 on the far end side, R (n), R (n−1),..., R (n−L + n) are set as level determination results for each M sample. If there are fluctuations with regularity such as negative, positive, negative,..., This can be regarded as a repetitive waveform, that is, a narrowband signal. The narrowband determination may be performed by calculating the variance of R (n) and determining that a narrowband signal is generated when the variance is greater than a certain threshold. This variance can be obtained by the following equation (17). In the equation (17), the average value of R (n) is MEAN_R (n).

MEN_R (n + 1) =
(1 · L) {MEAN_R (n) −R (n− (L−1)) + R (n + 1)} (17)
Next, the determination protector 190 will be described.

A propagation delay time estimator employed in a telephone network line echo canceller must estimate a delay time by using a far-end signal during communication as a test signal. For this reason, it is impossible to make a complete significance determination only by the above-described significance determiner 180 due to the influence of non-stationarity of the speaker signal itself, environmental noise around the speaker, double talk, and the like. Therefore, the propagation delay time estimator 10, provided with a determination protector 190, to evaluate the statistical reliability of the tau _min given from significance determiner 180. In the judgment protector 190, for example, majority logic may be applied as a practical and simple statistical reliability evaluation.

  FIG. 7 is an explanatory diagram showing the state transition of the first method and the second method in the case where the judgment protector 190 evaluates the statistical reliability based on the majority logic.

  FIG. 7A is an explanatory diagram showing an example of state transition in the first method in which the statistical reliability is evaluated based on the majority logic in the decision protector 190.

In the state transition diagram of FIG. 7A, first, after initializing various variables, it waits for the delay time τ _{min to} be reported from the significance determiner 180 (S100). Then, when τ _min is reported from the significance determiner 180, it is stored in the variable Delay, and the next report is awaited (S101).

In the state of step S101 described above, when τ _min is reported from the significance determiner 180, the values of Delay and τ _min are compared, and if they are the same, the process proceeds to the next step (step S102 described later). In FIG. 7A, “Delay = τ _min” is established, and “Yes” is indicated when the process proceeds to the next step.

In the state of step S101 described above, when the value of τ _min and Delay reported from the significance determiner 180 are different, the operation returns to the previous step (step S100 described above). In FIG. 7A, “Delay = τ _min” is not satisfied, and “No” is indicated when returning to the previous step.

Thereafter, the process proceeds in the same manner, and when reaching step Sm ₁ (“m _1” is a predetermined value set in advance), it is assumed that statistical reliability has been obtained, and the pseudo delay unit 12 is echoed. At the same time that the estimated propagation delay time to the path is reported, the operation returns to step S100 described above, and thereafter the above-described processing is repeated.

  FIG. 7B is an explanatory diagram showing an example of state transition in the second method in which the judgment protector 190 evaluates statistical reliability based on the majority logic.

In FIG. 7B, the decision protector 190 first waits for the delay time τ _{min to} be reported from the significance determiner 180 after initializing various variables (S200). Then, when τ _min is reported from the significance determiner 180, it is stored in the variable Delay, and the next report is awaited (S201).

When τ _min is reported from the significance determiner 180 in the state of step S201 described above, the values of Delay and τ _min are compared, and if they are the same, the process proceeds to the next step (step S202 described later). In FIG. 7B, “Delay = τ _min” is established, and “Yes” is indicated when the process proceeds to the next step.

In the state of step S201 described above, when the values of τ _min and Delay reported from the significance determiner 180 are different, the operation returns to the first step (step S200 described above). In FIG. 7B, “Delay = τ _min” is not established, and “No” is indicated when returning to the first step.

In the same manner, the process proceeds in the same manner. When reaching step Sm ₂ (“m _2” is a predetermined value set in advance), it is assumed that statistical reliability has been obtained, and an echo is sent to the pseudo delay unit 12. The estimated propagation delay time to the path is reported, and at the same time, the operation starts from step S200 described above, and thereafter the above-described processing is repeated.

In the two types of protection logic shown in FIG. 7A and FIG. 7B described above, the condition of Yes is Delay = τ _min , that is, Delay and τ _min reported this time are the same. However, even if they are not the same, they may be regarded as the same if the difference is less than or equal to the threshold value. For example, when the actual delay time, and tau _min, is an intermediate value of a value obtained by adding 1 to τ _min (τ _min +1) is significance determiner 180 estimates a delay report to determine protector 190 time values of the tau _min +1 of tau _min is, for example, may appear at 50% frequency. In this case, the Yes condition may be set to Delay ≦ τ _min ≦ Delay + 1.

  Next, the pseudo delay device 12 will be described. The pseudo delay device 12 is required to be sent back to the echo path 13 via the two-line transmission path 14 and return to the line echo canceller 11 via the two-line transmission path 15 via the echo path 13. After giving a delay time corresponding to the propagation delay time to the far-end signal, the signal is provided to the line echo canceller 11.

  Next, the pseudo delay device 12 will be described. The pseudo delay device 12 may be realized by using an existing shift register, or may be realized by a logical shift register by creating a circular queue in a general-purpose memory space. The delay amount set in the pseudo delay device 12 is assumed to follow the estimated value given from the propagation delay time estimator 10. However, since the estimated value of the propagation delay time estimator 10 includes an error corresponding to the time resolution, that is, the sampling interval of the envelope signal, the delay time created by the pseudo delay device is the propagation delay time estimator 10. The delay time may be set to be shorter by at least one unit time resolution than the value reported by.

(A-2) Operation of Embodiment Next, the operation of the propagation delay time estimator of this embodiment (propagation delay time estimation method of the embodiment) and the telephone according to the first embodiment having the above-described configuration. The overall operation of the communication system will be described.

  First, when a far-end signal is given to the decoder 111 via the two-wire transmission line 14, it is converted into a linear code applicable to addition / subtraction / multiplication / division calculation, and the converted one is given to the absolute value converter 121. On the other hand, an echo is given to the decoder 112 via the two-wire transmission line 15, and the echo is similarly converted into a linear code and given to the absolute value converter 122.

  When the converted signals are supplied from the decoders 111 and 112, the absolute value converters 121 and 122 convert the signals into absolute values and supply them to the low-pass filters 131 and 132.

  Next, when absolute value signals are given from the absolute value digitizers 121 and 122, the low frequency filters 131 and 132 block the high frequency components of the given signals, and the decimation devices 141 and 142 are cut off. Given to.

  When the signal from which the low-frequency components are cut off is applied from the low-pass filters 131 and 132 to the thinning-out devices 141 and 142, the thinning-out devices 141 and 142 reduce some of the given signals. Then, a signal with a reduced sampling frequency is generated and supplied to the level discriminators 151 and 152.

  When a signal with a reduced sampling frequency is given from the thinning-out devices 141 and 142, the level discriminators 151 and 152 perform level judgment and give positive or negative judgment results to the temporary storage units 161 and 162. It is done.

  When a signal whose sampling frequency is lowered is given from the thinning-out devices 141 and 142, the temporary storage units 161 and 162 store information related to the signals for a certain period of time, and the stored information (signal) is the least square. This is given to the arithmetic unit 170.

  When the information (signals) stored in the temporary storage units 161 and 162 is supplied to the least squares calculator 170, the near-end signal envelope signal supplied from the temporary storage unit 161 and the temporary storage unit 162 are provided. The square difference between the far-end signal and the envelope signal is calculated using the above-described equation (16) or the like using the time difference as a variable, and the calculated square difference is provided to the significance determiner 180, respectively.

When the cross-correlation coefficient calculated by the least squares calculator 170 is given to the significance determiner 180, the significance determiner 180 determines the significance of the time difference (τ _min ) when the square difference is minimized. inspection is performed on the first to fourth factors mentioned only tau _min it is determined that the significant is given to the determination protector 190.

When τ _min is given from the significance determiner 180, the determination protector 190 evaluates the statistical reliability of the τ _min based on the above-described flowchart (majority logic) shown in FIG. Only when it is determined that the pseudo delay unit 12 has the characteristic, the pseudo delay unit 12 is notified of the τ _min .

Then, in the pseudo delay device 12, a delay time corresponding to the time indicated by τ _min reported from the propagation delay time estimator 10 (determination protector 190) is given to the far-end signal, and the signal is sent to the line echo canceller 11 Provided to. The line echo canceller 11 generates a pseudo echo using an adaptive filter or the like based on the far-end signal given through the pseudo delay device 12, and removes the echo using the generated pseudo echo. .

(A-3) Effects of Embodiment According to this embodiment, the following effects can be achieved.

(A-1-1) In this embodiment, an envelope signal whose behavior is slower than that of the original audio signal is extracted from the audio signal, and the signal is thinned out by the decimation devices 141 and 142, thereby sampling. The speed can be reduced as compared with the signal before extracting the envelope signal, and the processing amount and memory capacity of the correlation calculation required for estimating the propagation delay time of the echo path can be reduced. For example, assuming that the effective sampling frequency of an envelope signal of a telephone voice signal with a sampling frequency of 8 kHz is a maximum of 200 Hz, 40 envelope signals can be thinned out into one envelope signal. As described above, in this embodiment, by extracting and thinning out the envelope signal from the audio signal, the sampling speed is reduced to about one tenth of the audio signal, and the correlation calculation required for estimating the propagation delay time of the echo path The processing amount and the memory capacity can be reduced to about several tenths compared with the case where the processing is performed using the audio signal itself.

(A-1-2) In this embodiment, propagation delay time estimation based on the envelope signal level of the audio signal is performed. The propagation delay time estimation based on the envelope signal level can be calculated by, for example, absolute value calculation and addition, as shown in the above equation (16), so that it is calculated in comparison with cross-correlation calculation that performs square root processing or division. The amount can be reduced.

(A-1-3) Since the level discriminators 151 and 152 calculate the level discrimination result of the voice envelope signal, the signal energy can be limited to positive and negative binary values. Therefore, the least squares calculator 170 can omit the normalization process and obtain the correlation between the two speech waveforms. In addition, since the least squares calculator 170 has only to perform an operation related to a binary code by converting to a binary code, the operation is performed in comparison with a case where an operation is performed on a signal before conversion to a binary code. The amount of processing and the memory capacity can be reduced.

(A-1-4) When the significance determination unit 180 passes all or part of the significance tests related to the above first to fourth factors, the calculation result by the least squares calculator 170 is regarded as significant. Since τ _min corresponding to the propagation delay time of the echo path 13 at this time is given to the determination protector 190, the frequency of estimating the erroneous time as the delay time can be reduced due to a malfunction. For example, by monitoring the near-end speaker envelope power and the far-end speaker envelope power, the correlation calculation is significant only when only the far-end speaker is busy and the echo is returning. Can be considered. In addition, in order to avoid malfunctions in the double talk state where the detection is incomplete by simply monitoring the near-end speaker envelope power and the far-end speaker envelope power, a threshold is set for the least square sum of the level discrimination results. Only the least square sum exceeding the threshold is considered significant. Further, a method for detecting the variance of the level discrimination results by the level discriminators 151 and 152 is provided, and when the variance of the level discrimination results is larger than the threshold, the calculation result by the least squares operator 170 is invalidated, for example, This prevents malfunctions caused by superimposition of low-frequency noise such as air-conditioning equipment on the voice envelope signal.

(A-1-5) For a malfunction that cannot be protected only by a short-term significance evaluation by the significance determiner 180, the decision protector 190 provides long-term majority logic protection for the significance evaluation. This increases the reliability of the estimation results.

(A-1-6) By estimating the propagation delay time to the echo path even during a call, even if the echo path is switched to another echo path during a call, such as a transfer service, a new eco Propagation delay time up to the path can be estimated, and by resetting the delay amount in the pseudo delay device, the echo cancellation operation in the line echo canceller can be maintained, thereby minimizing the degradation of call quality. can do.

(B) Other Embodiments The present invention is not limited to the above-described embodiments, and may include modified embodiments as exemplified below.

(B-1) In the above embodiment, an example in which the propagation delay time estimator of the present invention is applied to an echo canceller that cancels line echo in a telephone line has been described, but acoustic echo (for example, a speaker in a conference system or the like) It may be applied to an echo canceller that cancels echo generated by acoustic coupling between the microphone and the microphone.

 (B-2) In the above embodiment, the example in which the propagation delay time estimator is applied to the echo canceller has been described. However, the propagation delay time until the probe signal is emitted and reflected by the target is returned is measured. It can also be applied to systems that perform such operations. For example, such a system includes a remote search system using a radar or a sonar.

(B-3) Although the propagation delay time estimator and the echo canceller have been described as separate devices in the above embodiment, the propagation delay time estimator may be mounted on the echo canceller itself.

(B-4) In the above embodiment, the envelope generators of the far-end signal and the near-end signal are generated using the absolute value converters 121 and 122, the low-pass filters 131 and 132, and the envelopes thereof. For the line signal, the result of level discrimination by the level discriminator is subjected to calculation by the least squares calculator 170 or significance determination by the significance determiner 180, but the original signal is not the envelope signal. The results of level discrimination by the level discriminators 151 and 152 may be used for correlation coefficient calculation and significance determination.

  For example, the absolute value converters 121 and 122, the low-pass filters 131 and 132, and the decimation units 141 and 142 shown in FIG. 1 are omitted, and the signals output from the decoders 111 and 112 are subjected to level discrimination as they are. The data may be input to the devices 151 and 152.

  Further, in the above-described embodiment, the significance determination by the significance determiner 180 and the determination protection by the determination protector 190 are performed on the calculation result of the least squares calculator 170, but one configuration is omitted. Thus, it may be outputted as the estimation result in the propagation delay time estimator 10, or both may be omitted, and the calculation result of the least squares calculator 170 may be output as it is as the estimation result in the propagation delay time estimator 10. Anyway.

(B-5) In each of the above-described embodiments, by extracting the far-end signal and the near-end signal envelope signals using an absolute value converter, a low-pass filter, or the like, a lower frequency band can be obtained. Although the signal is generated, the configuration for generating a signal of a lower frequency band is not limited to extracting the envelope signal. For example, a signal in a lower frequency band may be generated by extracting a part of the frequency band of the far-end signal and the near-end signal using a band-pass filter and thinned by a thinning-out device.

(B-6) In the above embodiment, the level discriminators 151 and 152, as shown in FIG. 2 and FIG. 3 described above, convert the signal value of the input signal into a binary code based on the threshold, that is, The quantized bit is converted into a code represented by 1 bit. However, for example, it may be a code of three or more values (quantized bit is 2 bits or more), and fewer quantized bits than the input signal. The content of conversion is not limited as long as it is expressed by a number.

  For example, when converting to a four-value code (quantization bit number is 2 bits) represented by any one of 0, 1, 2, and 3, three thresholds are provided (“first threshold <second The threshold value <the third threshold value ”),“ 0 ”if the input signal value is less than the first threshold value, and“ 0 ”if it is less than the first threshold value and less than the second threshold value. The conversion may be made such that “1” is “2” if it is greater than or equal to the second threshold and less than the third threshold, and “3” if it is greater than or equal to the third threshold. Further, all or a part of the plurality of threshold values may have hysteresis characteristics as shown in FIG.

(B-7) The propagation delay time estimation in the above embodiment is calculated by the absolute value calculation and addition as shown in, for example, the above equation (16) using the least squares calculator 170. For example, the cross-correlation coefficient may be calculated using the time difference as a variable, and the time difference when the cross-correlation coefficient is maximized may be given to the significance determination unit 180 as an estimated value of the propagation delay time. The calculation method for calculating the estimated value of the delay time is not limited.

It is a block diagram which shows the detailed structure of the propagation delay time estimator which concerns on embodiment, and the positioning in a telephone communication system. In the level discriminator which concerns on embodiment, it is explanatory drawing shown about the example in the case of performing level determination simply on the basis of a threshold value. It is explanatory drawing shown about the example which performs the level determination using a hysteresis characteristic in the level discriminating device which concerns on embodiment. It is explanatory drawing shown about the structural example of the temporary memory part of the near end side which concerns on embodiment. It is explanatory drawing which showed the time relationship of the level discrimination | determination result by the level discriminator given to the temporary memory part which concerns on embodiment. It is explanatory drawing shown about the structural example of the temporary memory part by the side of a far end which concerns on embodiment. In the determination protector which concerns on embodiment, it is explanatory drawing shown about the state transition of the 1st system and 2nd system in case statistical reliability evaluation is performed based on majority logic. It is the block diagram which showed the example of arrangement | positioning of the conventional line echo canceller with a long propagation delay time to an echo generation point. It is explanatory drawing which showed the example of the impulse response in the conventional line echo canceller with a long propagation delay time to an echo reflection point.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Telephone communication system, 10 ... Propagation delay time estimator, 111, 112 ... Decoder, 121, 122 ..., Absolute value converter, 131, 132 ... Low-pass filter, 141, 142 ... Decimator, 151, 152 ... level discriminator, 161, 162 ... temporary storage unit, 170 ... least squares calculator, 180 ... significance judgment unit, 190 ... judgment protector, 11 ... line echo canceller, 12 ... pseudo delay unit, 13 ... Echo path, 15, 14 ... 2-wire transmission path.

Claims (7)

In a propagation delay time estimator that estimates a propagation delay time that is a time difference between a first discrete time signal and a second discrete time signal that were originally the same signal,
The first discrete time signal is converted into a third discrete time signal expressed with a smaller number of quantization bits, and the second discrete time signal is expressed with a smaller number of quantization bits. A quantized bit number converting means for converting to a fourth discrete time signal;
A propagation delay time estimator comprising delay time calculation means for calculating an estimated value of the propagation delay time using the third discrete time signal and the fourth discrete time signal.   The quantization bit number conversion means replaces each signal value of the first discrete time signal with either the first value or the second value based on the comparison result with the first threshold value. To convert the signal value of the second discrete time signal to the first value or the second value based on the comparison result with the second threshold value. The propagation delay time estimator according to claim 1, wherein the propagation delay time estimator is converted into the fourth discrete time signal by replacing any one of the signals. The first threshold has a hysteresis characteristic with respect to a change in the first discrete time signal,
The propagation delay time estimator according to claim 2, wherein the second threshold value has a hysteresis characteristic with respect to a change in the second discrete time signal.   The delay time calculating means calculates a time difference that minimizes a square difference between the third discrete time signal and the fourth discrete time signal as an estimated value of the propagation delay time. Item 4. The propagation delay time estimator according to any one of Items 1 to 3. A computer mounted on a propagation delay time estimator that estimates a propagation delay time that is a time difference between the first discrete time signal and the second discrete time signal that were originally the same signal,
The first discrete time signal is converted into a third discrete time signal expressed with a smaller number of quantization bits, and the second discrete time signal is expressed with a smaller number of quantization bits. A quantized bit number converting means for converting to a fourth discrete time signal;
Propagation delay time estimation using the third discrete time signal and the fourth discrete time signal to function as delay time calculation means for calculating an estimate of the propagation delay time program. In a propagation delay time estimation method for estimating a propagation delay time that is a time difference between a first discrete time signal and a second discrete time signal that were originally the same signal,
Quantization bit number conversion means, delay time calculation means,
The quantization bit number conversion means converts the first discrete time signal into a third discrete time signal expressed by a smaller number of quantization bits, and converts the second discrete time signal to Converting to a fourth discrete time signal represented by a smaller number of quantization bits;
The delay time calculating means includes calculating an estimated value of the propagation delay time using the third discrete time signal and the fourth discrete time signal. Estimation method. Equipped with the propagation delay time estimator according to claim 1,
An echo characterized in that the propagation delay time estimator is used to estimate a propagation delay time that is a time difference between a signal sent to the echo path and the echo, and the echo is canceled using the estimated propagation delay time. Canceller.

Images