Patent Translate Powered by EPO and Google Notice This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate, complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or financial decisions, should not be based on machine-translation output. DESCRIPTION JP2002135170 [0001] BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an echo cancellation apparatus, echo cancellation method, and recording medium recording echo cancellation program for erasing echoes which cause feedback and acoustic feedback mainly in loud speech systems. . [0002] 2. Description of the Related Art With the spread of voice conferences, there is a demand for the provision of a speech communication system which is excellent in simultaneous call performance and has little echo. An echo canceler is one that meets this requirement. FIG. 1 shows a block diagram of a conventional echo canceler. A / D converter 8 receives an incoming call in a speech communication system including a reception system from reception signal end 1 receiving reception signal x (t) to speaker 2 and a transmission system from microphone 3 to transmission output end 4 The signal x (t) is sampled, and the received signal x (t) is supplied to the pseudo echo path (FIR filter) 7, and the pseudo echo signal y ^ (k) from the pseudo echo path 7 is converted to A / D. The echo signal y (k) is eliminated by subtracting the echo signal y (k) sampled by the converter 5 from the echo signal y (k) by the subtracter 9 to obtain a residual signal e (k). [0003] 15-04-2019 1 Here, the pseudo echo path 7 needs to follow the time-dependent fluctuation of the echo path, and is estimated sequentially by the estimation circuit 6 so that the residual signal e (k) = y (k) −y ^ (k) approaches 0. By correcting the pseudo echo path 7, appropriate echo cancellation is always performed. For example, a learning identification method is used for successive correction of the impulse response h ^ (k) of the pseudo echo path 7, and the estimation algorithm is updated using the update equation (2) so as to satisfy the equation (1). Ru. y (k) = xT (k) h ^ (k + 1) (1) h ^ (k + 1) = h ^ (k) + β (k) x (k) (2) where β (k ) Is determined as follows. [0004] y (k) = xT (k) {h ^ (k) + β (k) x (k)} y (k) = xT (k) h ^ (k) + xT (k) β (k) x (k) y (k) -xT (k) h ^ (k) = β (k) xT (k) x (k) where e (k) = y (k) -xT (k) h ^ (k) and Then, e (k) = β (k) × T (k) × (k) β (k) = e (k) / xT (k) × (k) is substituted into equation (2), and finally the step size Using α, h ^ (k + 1) = h ^ (k) + αe (k) x (k) / xT (k) x (k) h ^ (k) = [h1 ^ (k), h2 ^ (k),..., hN ^ (k)] Tx ^ (k) = [x (k), x (k-1), ..., x (k-L + 1)] Te (k) = Y (k)-x T (k) h ^ (k) (α: step size (scalar amount, constant), L: number of taps, T: vector transpose), impulse response h ^ of pseudo echo path 7 Let k) approach the impulse response h (k) of the true echo path. [0005] Here, the impulse response of the echo path to be estimated is exponentially attenuated as shown in the reverberation curve of FIG. 2, and an exponential weighting algorithm is proposed to reflect it on the step size and to accelerate the convergence speed (for example, , JP-A-1220530). This method uses a step size matrix A instead of the scalar step size. A is a diagonal matrix, and its elements are [α1, α2, ..., αL]. Here, αi is represented by αi = α1γi-1 (3) (γ: exponential attenuation coefficient). This improves the convergence speed as shown in FIG. In FIG. 3, “ES (Exponential Weighted Step Size) method” indicates an exponential weighting algorithm, and “NLMS (Normalize Least Mean Square)” indicates a learning identification method. [0006] The estimation algorithm using the step size matrix A is updated using the update equation (5) so as to satisfy the equation (4). y (k) = xT (k) h ^ (k + 1) (4) h ^ (k + 1) = h ^ (k) + β (k) Ax (k) (5) 15-04-2019 2 where, β (k) Is determined as follows. y (k) = xT (k) {h ^ (k) + β (k) Ax (k)} y (k) = xT (k) h ^ (k) + xT (k) β (k) Ax (k) y (k) -xT (k) h ^ (k) = β (k) xT (k) Ax (k) e (k) = β (k) xT (k) Ax (k) β (k) = e Substituting in (k) / xT (k) Ax (k) equation (5), introducing a step size matrix, the impulse response of the pseudo echo path can be obtained by the following equation. h ^ (k + 1) = h ^ (k) + αe (k) Ax (k) / xT (k) Ax (k) [0007] The step size matrix simulating the impulse response of the echo path of the conventional echo canceler is usually set in a standard room. However, the impulse response of the echo path differs from room to room. FIG. 4 shows reverberation curves of impulse responses of two different rooms (room A and room B). These impulse response exponential attenuation amounts are different, and there is a problem that the conventional echo canceler can not estimate the impulse response with sufficient accuracy and at high speed and can not cancel the echo with sufficient performance. [0008] The present invention has been made in view of the above problems, and it is an object of the present invention to set the step size matrix to an optimum value and further accelerate the convergence speed by using the impulse response coefficient of the echo path after convergence. Do. [0009] SUMMARY OF THE INVENTION The present invention focuses on the fact that the impulse response of the pseudo echo path is equal to or very similar to the impulse response of the echo path when the echo signal can be eliminated by the pseudo echo signal. And estimating the exponential decay coefficient γ of the step size matrix that is optimal for the impulse response of the echo path from the coefficients of the impulse response of the pseudo echo path at that time. According to the present invention, as described above, since the exponential attenuation coefficient of the step size matrix can be set to the impulse response of the echo path, an echo canceler with good convergence characteristics can be obtained. 15-04-2019 3 [0010] The invention according to claim 1 generates a pseudo echo path by sequentially estimating the impulse response of the echo path by the estimation algorithm from the transmission signal to the echo path and the echo signal after passing through the echo path of the transmission signal. And an echo canceler for eliminating the echo signal by subtracting the pseudo echo signal obtained by using the transmission signal as the input of the pseudo echo path from the echo signal, the step size in the estimation algorithm being the pseudo echo A step size matrix storage circuit for storing a step size matrix weighted by the exponential decay characteristic of the impulse response of the sound field for setting each coefficient of the impulse response of the path, the pseudo acoustic signal from the echo signal And a convergence determination circuit that determines whether or not the echo signal can be satisfactorily eliminated based on the residual signal from which the And having a damping coefficient calculation circuit to reflect the better the decision that can be erased by calculating the damping coefficient from the coefficient of the estimated echo path Motoma' from the estimation algorithm the step size matrix No.. [0011] The invention according to claim 2 is characterized in that, in the echo canceler according to claim 1, the estimation algorithm is any one of a learning identification method, an LMS method, an affine projection method, or an RLS method. The invention according to claim 3 generates a pseudo echo path by sequentially estimating the impulse response of the echo path by the estimation algorithm from the transmission signal to the echo path and the echo signal after passing through the echo path of the transmission signal. And an echo cancellation method for canceling the echo signal by subtracting the pseudo echo signal obtained by using the transmission signal as the input of the pseudo echo path from the echo signal, the step size in the estimation algorithm being the pseudo echo A step size matrix weighted by the exponential decay characteristics of the impulse response of the sound field to be set for each coefficient of the impulse response of the path is stored, and based on the residual signal obtained by subtracting the pseudo acoustic signal from the echo signal. The estimation algorithm is determined by determining whether the echo signal can be erased well and the echo signal can be erased well. Wherein the calculating the attenuation coefficients from the coefficient of al Motoma' pseudo echo path is reflected in the step size matrix. [0012] 15-04-2019 4 The invention according to claim 4 is characterized in that, in the echo cancellation method according to claim 3, the estimation algorithm is any one of a learning identification method, an LMS method, an affine projection method, or an RLS method. The invention according to claim 5 generates a pseudo echo path by sequentially estimating the impulse response of the echo path by the estimation algorithm from the transmission signal to the echo path and the echo signal after passing through the echo path of the transmission signal. Recording an echo cancellation program for causing a computer to execute the process of erasing the echo signal by subtracting the pseudo echo signal obtained by setting the transmission signal as the input of the pseudo echo path from the echo signal Storing in the medium, a step size matrix weighted by an exponential decay characteristic of an impulse response of a sound field for setting a step size in the estimation algorithm for each coefficient of an impulse response of the pseudo echo path; Whether the echo signal has been successfully eliminated based on the residual signal obtained by subtracting the pseudo acoustic signal from the echo signal And a process of calculating an attenuation coefficient from the coefficient of the pseudo echo path obtained from the estimation algorithm based on the judgment that the echo signal can be erased well and reflecting it on the step size matrix. . [0013] The invention according to claim 6 is characterized in that, in the recording medium recording the echo cancellation program according to claim 5, the estimation algorithm is any one of a learning identification method, LMS method, affine projection method, or RLS method. . [0014] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 5 shows an embodiment of the present invention, and shows the configuration of a portion corresponding to the estimation circuit 6 in FIG. 11The convergence judgment circuit shown by, observes the residual signal e (k) and becomes sufficiently small (becomes smaller than the set threshold) or the received signal x (k) and the residual signal e (k) It is observed that the received signal x (k) is sufficiently large and the residual signal e (k) is sufficiently small, that is, it is determined whether the echo signal can be satisfactorily eliminated (the adaptive filter has converged or not). I do. As an example of the judgment that the adaptive filter has converged, the ratio of the power of the residual signal e (k) to the power of the received signal x (k) is smaller than a certain threshold, that is, the echo is good by the adaptive filter When it is considered to be erased, the information (signal) that echo 15-04-2019 5 cancellation is possible is transmitted to the attenuation coefficient calculation circuit shown at 12. The attenuation coefficient calculation circuit receives information on the coefficient ^ of the pseudo echo path estimated from the estimation circuit based on the signal that echo cancellation can be performed, and reflects it to calculate the step size matrix A, and the calculated step size Pass the matrix to the estimation circuit. The estimation circuit can estimate the impulse response at high speed by estimating the impulse response using the calculated step size matrix. [0015] In the damping coefficient calculation circuit 12 of FIG. 5, the step size matrix A obtains the reverberation time using the estimated impulse response (adaptive filter coefficient) of the echo path, and the elements of each step size matrix are obtained from the equation (3) above. Ask for The exponential attenuation coefficient γ for determining each element of the step size matrix is determined, for example, as follows. Here, an example is shown in which the reverberation time is determined from three observation points. Assuming that the adaptive filter length is L, attenuation amounts at three observation points (L / 4, L / 2, 3L / 4) are obtained as follows. [0017] The inclination is obtained from the obtained R1, R2 and R3, and the average inclination Δg of the reverberation curve per one tap is calculated. Δg = 2 {(R3−R2) + (R2−R1)} / L reverberation time (time until sound energy attenuates from the first steady state by 60 dB) Rt can be calculated by the following equation. Rt = {(− 60−R2) / Δg + L / 2} / fs where fs is a sampling frequency. The exponential decay coefficient γ of the equation (3) is given by the following equation from the obtained Rt. [0018] γ = exp (−μ) where μ (index representing the attenuation characteristic) is determined by the following. 10log 10 exp (-μ Rt) 2 = -60 log 10 exp (-μ Rt) = -3 exp (-μ Rt) = 10 -3-μ Rt = log e 10 -3 μ = log e 10 -3 / R t Therefore, γ in equation (3) is It can be asked. γ = exp (-3 loge 10 / Rt) 15-04-2019 6 [0019] FIG. 6 shows the simulation result according to the present invention, and it can be seen that, after the adaptive filter has converged, the impulse response of the echo path changes most rapidly when it changes. Referring to FIG. 6, in the initial stage of learning (before echo path change), the same convergence speed as that of the conventional ES method is shown, but the threshold considered to be able to sufficiently reduce the echo (for example, the echo reduction amount is 25 dB Above, etc.) The amount of attenuation is determined from the impulse response estimated in the following cases, and is reflected in the step size. Since the step size is used for impulse response estimation after the echo path changes, the convergence speed is faster than the conventional ES method. [0020] As the estimation algorithm used in the present invention, in addition to the learning identification method, LMS (Least Mean Square) method, affine projection method, RLS (Recursive Least Square) method can also be used. The echo canceler according to the present invention comprises a computer having a CPU, a memory, etc., a user terminal as an access subject, and a recording medium, and the recording medium is a machine-readable record such as a CD-ROM, a magnetic disk, or a semiconductor memory. A medium, the echo canceler program recorded therein is read by the computer to control the operation of the computer to realize the above-described components on the computer. [0021] According to the present invention, by adaptively updating the exponent weighting algorithm of the echo canceler, the convergence speed can be increased when the impulse response of the echo path changes. In this case, it is not necessary to estimate the reverberation time by the initial training. 15-04-2019 7

1/--страниц