JPS5964932A - Band division type echo erasing device - Google Patents

Abstract

PURPOSE:To erase echoes with high efficiency, by dividing signals into plural bands and giving a low-band converting process when necessary to the signal of each band to lower the sampling frequency. CONSTITUTION:The received input signals are divided into (n) units of frequency band components by a band dividing filter group 12. The lowest band signal is fed directly to a partial echo canceller 16, and other high band components pass through a low-band converting circuit group 13 and then fed to partial echo cancellers 14, 15, etc. Then the signal of each band is returned to the signal of the original band through a high band converting circuit group 17 and a band synthesizing filter group 18 and then fed to an echo circuit 8 in the form of a received output signal. The signals returned to the circuit 8 are separated to the component of each band by the filter group 12 having the same constitution as the reception side. The high band component passes through the circuit group 13, and then a difference between the high band and a pseudo echo output is obtained by a subtractor circuit 10 within the canceller 16.

Description

[Detailed Description of the Invention] (Technical Field) The present invention relates to an echo cancellation device that is used when the target echo signal has a wide band and the impulse response length of the echo path is large. .

(Background Art) Conventional devices of this type have a relatively narrow band (<4 to 11z
), the echo canceller is applied only for a short period of time compared to the impulse response length of the system for high frequency components, and long echoes are applied only for low frequency components. A method was used to ensure erasing time.

Figure 1 shows the block configuration of the conventional equipment in this building (1984 Institute of Electronics and Communication Engineers National Conference Lecture No. 1994).

In the figure, 1 is the input terminal for the received signal, 2 is the low-pass filter, 3 is the high-pass filter, 4 is the low-pass echo canceller, 5 is the adder, 6 is the echo canceler for the entire range,
7 is an output terminal for a received signal, 8 is an echo path, 9 is an input terminal for a transmission signal, 10 is an output terminal, and 11 is an output terminal for a transmission 1g signal.

The received input signal is separated into high and low frequencies by a high-pass filter 2 and a low-pass filter 3, the high-frequency signal 1g goes directly to the adder 5, and the low-frequency signal goes to a low-frequency echo canceller. 40 inputs and one input of the adder 5. The output of the adder 5, ie, the received output signal, becomes an input to the echo canceller 60 for the entire area and is also input to the echo path 8.

The output signal of the echo path, that is, the transmission input signal, is subtracted from the pseudo echo output by a subtraction circuit 10 in the echo canceller, and is separated into high and low frequencies by a high-pass filter 2 and a low-pass filter 3. The high-frequency signal enters the adder 5 as it is, and the low-frequency signal becomes the other input to the adder 50 after being subtracted from the pseudo echo output. Adder 5
The output of becomes the transmission output signal.

The full-band echo canceller is a fixed-coefficient echo canceller with a number of taps that is much shorter than the impulse response duration of the system, and the coefficients are fixed by measuring the impulse response of the system by synchronous addition when the device is started up. . On the other hand, the low-frequency echo canceller is configured as an adaptive type.

The reasons for adopting these configurations are (1) that the spectrum of the audio signal is concentrated in the low range (:i) the lower the frequency of the received signal, the longer the response ( It is thought that such an assumption does not necessarily hold true, considering the fact that it tends to cause noise, and considering its application to a conference room where the audio band is widened and the acoustic processing is not sufficient. The same processing as for the low range is required for the high range.

Furthermore, configuring the full-band echo canceller with fixed coefficients has the disadvantage that a sufficient amount of echo cancellation cannot be achieved, considering the fluctuations in the echo path due to changes in indoor temperature and movement of people.

Furthermore, the processing of the two types of echo cancellers is completely different, requiring separate hardware and arithmetic algorithms, which complicates the configuration (there were 5 drawbacks).

(Problem to be solved by the invention) In the present invention, when canceling an echo, the input/output signal is divided into a plurality of bands, and the sampling frequency is lowered by performing low-frequency conversion processing on the signal of each band as necessary. The purpose of this method is to efficiently cancel echoes of threads whose echoes have a wide frequency band and whose echo path has a long impulse response length.

(Structure and operation of the invention) FIG. 2 is a basic configuration diagram of the present invention, in which 12 is a band division filter, 13 is a low-frequency conversion circuit group, 14 is a partial echo canceller for the highest frequency band, and 15 is a partial echo canceller for the highest frequency band. A partial echo canceller for the second frequency band from the lowest, 16 a partial echo canceller for the lowest frequency band, ] 7 a high frequency conversion circuit group, 18 a band synthesis filter group, 7 a receiver (
, 1 signal output terminal, 8 an echo path, 9 a transmission signal input terminal, 10 a subtraction circuit in the echo canceller,] 1
is the output terminal of the transmission signal.

Next, the operation of the system will be explained. The received input signal is divided into n frequency band components ((() by the band division filter group 2), the lowest band signal goes directly to the partial echo canceller 1G, and the other high frequency components are sent to the low frequency conversion circuit group 13.
After passing through, the signal is input to partial echo cancellers 14, 15, etc. After that, the signals of each band are transferred to the high frequency conversion circuit group 17.
, band synthesis filter group 18, the signal is returned to the original band signal, and is input to the echo path 8 as a received output signal.

The signal returned from the echo path 8, that is, the transmission input signal, is separated into components of each band by a group of band division filters 12 having the same configuration as that on the receiving side, and after passing through a group of low frequency conversion circuits 3 for high frequency components. , the difference from the pseudo echo output is taken by a subtraction circuit 10 in the echo canceller, and the output signal passes through a high frequency conversion circuit for high frequency components.
The low-frequency component enters the band synthesis filter group 18 as it is, and its output becomes the transmission signal.

Each partial echo canceller is an adaptive echo canceller, and echo cancellation is performed by the output signal (error signal) of the It-Ji device 10 within the echo canceller. The operating band of the partial echo canceller and the weighting of the amount of echo cancellation in each band should be optimized according to the characteristics of the input signal and echo path.
Becomes Noh.

Figure 3 shows an example of application to a conference system that connects multiple conference rooms with each other via transmission lines, and 19 is an analog
2o is a resampling circuit, 21 is a partial echo canceller, 22 is a digital-to-analog converter, 23 is a speaker driving amplifier, 24 is a speaker, 25
is a microphone, 26 is a microphone amplifier,
27 is a conference room, and 28 is a transmission line.

In general, such conference systems use a so-called public address system using a microphone and speaker as the audio system, so the echo path is the space in the conference room from the microphone to the speaker, so the impulse response length is 1100 [ +n sec].

Also, high quality audio (band 7 to 10 [KIIZ) level]
If you try to process the signal in its original band, the sampling frequency will be 171 to 20 [K1-
In a typical configuration, the number of taps of the transversal filter in the echo canceller is several thousand and 10,000 taps (the echo cancellation time is 250 [
m sec :l, sampling frequency 16 [K1
1z'], this would amount to 8,000 taps), which is almost impossible to realize from the viewpoint of hardware and arithmetic processing.

In the embodiments shown in FIGS. 2 and 3, the received signal input terminal 1
The signal reaching the received signal output terminal 7 is subjected to band division/low-band conversion and high-band conversion/band synthesis processing. However, such processing is not always necessary, and the received signal input terminal 1 and the received signal output terminal may be directly connected, and in that case, the high frequency conversion circuit 17, band synthesis filter 18, and tee Digital to analog converter 22 is eliminated.

From the viewpoint of noise characteristics, it is advantageous not to perform the above processing.

By applying the band division type echo cancellation device of the present invention, the number of taps in each partial echo canceller can be reduced to a few to about 1,000 taps (echo cancellation time of 250 [
m sec:), audio signal band 8 CI<1-1z
], number of band divisions 4, sampling frequency 4 CI<I
As (z), the number of taps of the partial echo canceller is 10.
00 taps) is fully achievable with current technology.

There are various band division methods that can be considered, but on the receiving side when the integer band sampling method is adopted, it is relatively easy to convert high-frequency components to low-frequency components, and good transmission characteristics can be obtained depending on the filter configuration. Figure 4 shows the spectrum of each part.

Figure 4 (al is the original received signal, (bl is 16 (K1-1
z) shows the spectrum of the signal after sampling (analog-to-digital conversion). Also, (C1, (d
l, (e), (f) are band division filter group 12
(C) corresponds to the lowest frequency signal.

In history, (4 [K1-1 by gl beam sample circuit 20
z: ] lowest frequency signal resampled (11) y
('), (J) are high-frequency signals or low-frequency conversion circuit group 13 (
In this case, the spectrum of the signal after passing through a simple resampling circuit with a sampling frequency of 4 [KHz] is shown.

Numbers 29 to 38 in FIG. 4 indicate corresponding positions in FIG. By appropriately designing the system, the sampling rate is reduced isometrically, and the echo canceller method can be applied even to systems with large impulse response lengths that include spatial propagation paths, such as conference rooms.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a conventional spatial propagation echo canceller, FIG. 2 is a basic diagram of the present invention, FIG. 3 is an example of the application of the present invention to a video conference room, and FIG. 4 is a diagram of the present invention. This is the spectrum of each part on the receiving side. 1... Input terminal for received signal, 2... High-pass filter, 3... Low-pass filter, 4... Low-pass echo canceller, 5... Adder, 6... Whole area echo canceller for use, 7...output terminal for received signal, 8...
・Echo path, 9...Input terminal for transmission signal, 10...
・Subtractor, 11... Output terminal for transmission signal, 12...
Band division filter group, 13... Low frequency conversion circuit group, 14... Partial echo canceller for the highest frequency band, 15... Partial echo canceler for the second frequency band from the lowest, 10 ...Partial echo canceller for the lowest frequency band, 17 City area conversion circuit output, )8...IiY area synthesis filter group, 15)...Analog-to-digital converter, 20
・Resample circuit, 21...partial echo canceller,
22・Digital/analog converter, 2:3・Speaker, drive amplifier, 24・Speaker, 2
5. Microphone, 26. Microphone amplification 8
S, 27 chamber, 28...transmission line, 20...
Received signal 1, 0. Received signal that has passed through the analog-to-digital converter, 31~:31 The received signal that has passed through the band division filter is on the low frequency side, 3:3 is on the high frequency side), ;35・Output of Lisa nozzle circuit, :36~,38 Output of low frequency conversion circuit. Patent applicant Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] It has an echo canceller that processes a signal obtained by sampling the received input signal to create a pseudo output signal of the echo path, and a subtracter that subtracts the output from the transmitted input signal, and remains in the transmitted output signal that is the output of the subtractor. An echo cancellation device that reduces echo components includes a band division filter that divides each of the received input signal and the transmitted input signal into a plurality of bands, and a low-frequency conversion process for the signal components of each band other than the lowest band. A low-frequency conversion circuit, a partial echo canceller that operates on each output of each low-frequency conversion circuit, a subtracter that subtracts the output from the transmission input signal for each band, and high-frequency conversion of the output of the subtractor. A band-splitting type echo canceling device characterized by having a high-frequency conversion circuit and a band synthesis filter for band-synthesizing and providing a transmission output signal.