JPS6210933A - Method for detecting double talk - Google Patents

Abstract

PURPOSE:To make detection high speed and high sensitivity by obtaining power of signals before removing echo by an echo canceler and power of signals after removing echo by integration and detecting double talk from the power ratio. CONSTITUTION:Sensitivity of detection of double talk is heightened by operating equations (5), (6), (7) using (ej) of large variation corresponding to change of n(t). Here, K is maximum value of power ratio of y(t) and n(t) when detection of double talk is necessary, and (q) is integer coefficient used in equations (6), (7) and both of them can be set to proper values. Power calculation of equations (6), (7) can be executed by two times of multiplication, subtraction, shifting and addition using necessary memories for two data. Thus, as power of (ej) and n(t) becomes nearly equal when an echo canceler is converged sufficiently, when the power of n(t) increases, (ej) also increases, and double talk can be detected quickly from the inequality (5).

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a double talk detection method in a system to which a digital echo canceler is applied to remove receiver echo caused by 2-wire to 4-wire conversion or the like.

(Prior Art) FIG. 1 shows the configuration of a system to which a digital echo canceller is applied. In Fig. 1, 1 is a digital signal receiving end from a 4-wire line, 2 is a ≠ converter, 3 is a hybrid circuit, 4 is a connecting end of a 2-wire line, 5 is a balancing network, 6 is a ψ converter, and 7 is a ψ converter. 8 is an echo canceller, and 9 is a digital signal transmission end to the 4-wire line.

Next, the operation of the above system will be explained. The digital signal at the digital signal receiving end 1 is called the far end input, and
Let it be expressed as j. This far end input is converted by the D/A converter 2 into an analog signal (1). x (
A part of the signal t) is input to the converter 6 as an echo, and this echo signal is expressed as y(t).

The path through which this echo passes is called the post-echo removal path, and its characteristic is defined as a quantity called echo noise, which is the power ratio of y (t) and x (t). Note that the echo path loss is expressed in AECHO [dB]. In addition, input signals from the two-line line are input to the φ converter 6 via the connection terminal 4 and the hybrid circuit 3. This signal is called a near-end input and is expressed as n(t).

Therefore, in the N raw converter 6, s (t) = y (t
) + n (t) analog signals are input and digital signals are output. This digital signal output will be represented by 8j. Further, the digital signal at the digital signal transmitting end 9 is called a residual echo signal and is expressed by ej.

The echo canceller 8 successively estimates the impulse response of the echonox using the residual echo signal ej and the far-end input column vector Xj=(xjxj10...Xj-N).

Then, at the next sample point, the estimated impulse response is
A pseudo echo is generated from the input column vector X j +1. At this time, the subtracter 7 removes the pseudo echo from sj+1 to obtain e j+1.

However, when estimation is continued in a double talk state with a large near-end input n(t), the estimated impulse response diverges to a value completely different from the actual impulse response, and conversely, the estimation error is added to 8j. I'm going to go to the middle of the day. Therefore,
It is equipped with a function that stops estimation when there is double talk. This double talk is detected by utilizing the fact that when a coil is used in the hybrid circuit, the echo path loss is AECHO≧6(1).

In this way, by performing the calculations of equations (2), (3), and (4), double talk can be detected for any echo/Js using the hybrid coil.

(Problems to be Solved by the Invention) However, the above method has a problem in that the divergence of the estimated impulse response cannot be sufficiently suppressed because the double talk detection sensitivity is low. Also, formula (3)
The power calculation in (4) also requires 2m data memories, and when the calculation is performed by software, there is a problem that it takes a lot of processing time.

The present invention is intended to solve these problems, and aims to provide an excellent double talk detection method that is high speed and highly sensitive, with a small amount of calculation and memory.

(Means for Solving the Problems) In order to achieve the above object, the present invention detects double talk by using ej that changes greatly in response to changes in n(1), S, = S, , +2-q(S,'-8,-1)
(6) pj pj J
The detection sensitivity is increased by calculating P. Here, K is A of y (t) and n (1) when double talk detection is required? The maximum value of the power ratio, q, is an integer coefficient used in equations (6) and (7), and both can be set to appropriate values. Also, formula (6) (
For the power calculation in 7), the required memory is 2 data and 2 data.
Can be performed by multiplying, subtracting, shifting, and adding times.

(Function) The present invention has the following effects based on the method described above. In other words, when the echo canceller is sufficiently converged, the powers of ej and n(1) are about the same, so n
When the noise in (1) increases, ej2 similarly increases, and according to equation (5), it is possible to quickly detect pull talk. In addition, since the change in the value of equation (5) with respect to the power change of n(1) is sufficiently large, integral calculations such as equations (6) and (7) can be applied to the nokwar calculation without significantly deteriorating the detection sensitivity. As a result, the memory and calculation amount required for double talk detection can be reduced.

(Example) Figure 2 shows the operation flow of an example of the present invention.
This is what I did. xj is a far-end input signal, and J is a digital signal obtained by φ-converting the weighted signal of the echo signal y (t) and the near-end input signal n(1). Also, xj is an input signal sequence vector, and Hj is an estimated impulse response at sample point j'. Sample value X.

When J Sj is input, the estimated impulse response H9
The residual echo e is calculated and output.

Further, the power of sj #ej is calculated and double talk judgment is performed. When double talk occurs, the step gain α is set to zero and estimation is stopped, but when double talk occurs, Hj is updated as α=2-3.

As described above, this embodiment has the advantage that the power of the near-end input n(t) enables double talk detection of -15 dB of the power of the echo signal y(t).

Further, when near-end input n(1) having the same power as y(t) is applied, double talk is detected with a delay of 4 samples, so there is an advantage that double talk can be detected at high speed.

In addition, in the example, K in equation (5) is set to 15, but
It may be increased within a range that does not interfere with the progress of estimation.

In this case, the effect of further increasing double talk detection sensitivity can be obtained. Furthermore, q in equations (6) and (7) is set to 5
However, it may be made smaller as long as it does not affect the safety of the transition from double talk to single talk. In this case, it is possible to obtain the effect of further increasing the double talk detection speed.

(Effects of the Invention) As is clear from the above embodiments, the present invention has the following effects.

(1) Since sj and ej are used for double talk detection, double talk detection can be performed at high speed and with high sensitivity.

(2) The threshold for double talk detection is y(
Since it is determined by the power ratio of 1) and n(1), a fixed threshold can be used for any echonoise.

(3) Near-end input: Since ej, which closely follows the changes in (1), is used, integral calculation can be applied to node calculation without significantly deteriorating detection sensitivity, and memory and calculation amount can be saved.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a system to which a digital echo canceller according to the present invention is applied, and FIG. 2 is a flowchart of an embodiment of the present invention. DESCRIPTION OF SYMBOLS 1...Digital signal receiving end, 2...D/A converter, 3...Hybrid circuit, 4...2-wire line connection end, 5...Balancing network, 6... Converter, 7... Subtractor, 8... Echo canceller,
9...Digital signal transmission end. Figure 1

Claims (1)

[Claims] In a system to which an echo canceller is applied, the power of a signal before echo cancellation by the echo canceller and the power of a signal after echo cancellation are determined by integral calculation, and double talk is detected based on the power ratio. Double talk detection method.