JPS6348210B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an echo suppressor for eliminating reverberation signals (echoes) caused by impedance mismatching of two-wire, four-wire converters, etc. in telephone lines.

A typical telephone line consists of a four-wire line and a two-wire line. Reflection is likely to occur due to impedance mismatch in a two-wire/four-wire converter (hybrid) that connects these two lines. On relatively short-distance telephone lines, this reflection rarely interferes with communication. However, in telephone lines that involve long signal propagation times, such as lines that include communication satellites, the above-mentioned reflections become echoes and cause major communication disturbances.

Echo suppressors or echo cancellers are used to address the echoes described above.

Both devices are inserted into a four-wire line. An echo suppressor compares the level of the signal on the receiving side with the level of the signal on the transmitting side, and only if the signal level on the receiving side is greater than a certain value and greater than the level of the transmitting side signal, a large signal is sent to the transmitting side. It prevents echoes occurring on two-line lines by providing loss (or disconnection). Echo suppressors can be constructed relatively easily and are already in commercial use for long-distance lines.

However, since the echo suppressor disconnects the line as described above, it has the disadvantage that partial disconnection of the conversation signal is unavoidable.

An echo canceller has been proposed as a method for eliminating such disconnection of speech signals. An echo canceller creates a filter that has the same characteristics as the echo path (the path from the output terminal on the receiving side of the echo canceller to the input terminal on the transmitting side of the echo canceller through the hybrid), and the input of this filter When the receiving side signal is added as , the same signal as the echo signal appearing at the transmitting side input terminal of the echo canceller is obtained at the output of the filter. By subtracting the output signal of this filter from the signal appearing at the transmitting input terminal, the echo signal can be canceled. This is the principle operation of an echo canceller, and it is an inherently superior method because it does not involve cutting off speech signals. However, since an extremely complicated filter must be constructed, an increase in hardware size is inevitable. For this reason, echo cancellers have not yet been put into practical use.

As mentioned above, several methods have been proposed to deal with the echo phenomenon, but none of them are simple in terms of hardware and have sufficiently good characteristics.

SUMMARY OF THE INVENTION An object of the present invention is to provide an echo suppression device that does not interrupt conversations, does not interfere with communication due to echoes, has high quality characteristics, has a simple circuit configuration, and can be miniaturized.

According to the present invention, there is an echo presser provided on the four-wire side of a two-wire, four-wire converter in a telephone line, which includes: (a) an analysis circuit for extracting the period T of a signal on the receiving side of the four-wire line; (b) a four-wire a filter including a delay circuit to which a transmission side signal of the line is supplied and whose delay amount is set to the T; and a subtraction circuit that outputs a difference signal between the transmission side signal and the output of the delay circuit; An echo suppressor is obtained that transmits the filter output to the far end subscriber.

A feature of the present invention is that echoes are removed by focusing on the periodicity of the audio signal.

Audio signals can be divided into voiced sounds (such as vowels) with relatively large amplitudes and unvoiced sounds with relatively small amplitudes.

The former mainly uses vibrations of the vocal cords as a sound source, and the latter uses air turbulence caused by the constriction of the vocal tract as a sound source. Therefore, voiced sounds generally have strong periodicity (pitch period), and unvoiced sounds have noise-like properties.

If there is a reflection on a long-distance telephone line, the voice signal becomes an echo and is heard. Naturally, a voice signal with a high level becomes a large echo signal and returns to the speaker, and a voice signal with a low level causes a small echo signal. Become.

Telephone lines always have some loss, and the reflections from impedance mismatches are not that great, so the echoes of small voice signals are less audible. As mentioned above, since unvoiced sounds generally have a small signal amplitude and voiced sounds have a large signal amplitude, it is safe to assume that only voiced sounds are mainly heard as echoes. Voiced sounds exhibit a fairly strong periodicity called pitch. Therefore, if a voiced sound is input to the received signal, the period of its pitch is measured, and if a filter is inserted on the transmitting side so that the gain of the signal at that period is zero, even if there is a reflection, it will be blocked by the filter. Therefore, no reverberation phenomenon occurs. Even if a near-end subscriber makes a sound when a received signal is present, the near-end subscriber's signal will easily pass through the filter because the respective pitch periods are different. Echo suppressors based on this type of operation do not perform switching on the transmitting side. Therefore, unlike conventional echo suppressors, the problem of cutting off the audio signal does not occur.

FIG. 1 is a block diagram of a first embodiment according to the present invention. The voice signal of the far-end subscriber passes through the receiving side input terminal 1, the receiving side output terminal 2 and the hybrid 20 of the echo suppressor 10, and is delivered to the near-end subscriber 30.
is supplied to the sender, and some leaks to the transmitter. In the echo suppressor 10, the receiving side signal is input to a pitch extraction circuit 120, where the pitch period of the receiving side signal is extracted.

The determined pitch period T is applied to the filter 110. The signal of the near-end subscriber 30 is sent to the far-end subscriber through the transmitting input terminal 3, the filter 110, and the transmitting output terminal 4.

FIG. 2 shows an example of the filter 110 of this example. The signal I input to the filter 110 is applied to a delay circuit 111 and a subtraction circuit 112. The delay circuit 111 is realized by a CCD or BBD, and its delay time is determined by the signal T supplied from the pitch extraction circuit 120. The output of delay circuit 111 is input to subtraction circuit 112. Subtraction circuit 112 performs subtraction of the two input signals, and the result becomes the output of filter 110. A filter configured as shown in FIG. 2 removes signals with a period T and signals with a period that is an integer fraction of T.

Expressed in another way, the gain at frequency 1/T and frequencies that are integral multiples thereof is zero. Therefore, this type of filter is called a comb filter.

According to this embodiment, even if the receiving side signal leaks to the transmitting side and becomes an echo signal, the echo signal is removed by the filter 110 because the periods of both signals are the same. On the other hand, since the period of the near-end subscriber's signal is different from the period of the receiving side signal, the near-end subscriber's signal easily passes through the filter.

As described above, in the configuration of this embodiment, when an echo suppressor is used, an echo phenomenon does not occur, and furthermore, conversational sounds are not cut off, so that extremely favorable conversation can be realized.

FIG. 3 shows an embodiment of a pitch extraction circuit. Various pitch extraction methods have been proposed. These examples are “A Comparative Performance Study of Several Performance Algorithms”.
Performance Study of Several Pitches
Detection Algorithms) [LR Rabiner et al., IE Transactions on Acoustics, Speech, and Signal Processing Vol. ASSP]
24, No. 5 (LRRabiner et.al., IEEE
Transactions on Acoustics, Speech, and
Signal Processing) Vol.ASSP―24, No.5,
(October 1976, pp. 399-418)]. Here, a general method suitable for real-time processing will be used as an example.

The received signal X is converted into a digital signal x _j at a sampling frequency of 8 kHz by the AD converter 121 at sampling time j. The output x _j of the AD converter 121 is applied to an autocorrelation circuit 122 . The autocorrelation circuit 122 has approximately 110 autocorrelation coefficients R(i)
Find (i = 14, 15, ..., 120). R(i) is given by R(i)= _M 〓 ^k=1 x _jik x _jk . M is 200. When a voiced sound is input, a waveform as shown in FIG. 4 is obtained by arranging the autocorrelation coefficients R(i) in the order of i. A strong peak appears in the autocorrelation coefficient of a periodic signal at the pitch period. In FIG. 4, the peak of the autocorrelation coefficient is obtained at i=T, so the period of the received signal X is T. This peak value detection process is executed by the maximum value circuit 123.

As described above, echo signals can be blocked by combining the filter 110 and the pitch extraction circuit 120.

However, this embodiment assumes that the audio signal is stationary (signal properties do not change over time). Therefore, although the filter 110 works effectively in the stationary part of a vowel, it cannot be said to be sufficient in blocking echoes in the beginning and end parts of a vowel.

FIG. 5 is a block diagram showing a second embodiment of the present invention. The difference between this embodiment and the first embodiment is that the pitch extraction circuit 120 outputs a control signal a in addition to the period T. In this embodiment, effective echo prevention is performed even when a non-stationary signal is received by calculating the degree to which the signal increases or attenuates. When the amplitude of the signal is increasing, when the autocorrelation coefficient is calculated, R(T) at one pitch away is R
(O). In the second embodiment, the division circuit 124 calculates the ratio a=R of the autocorrelation coefficients R(O) and R(T) determined by the autocorrelation circuit 122.
Outputs (O)/R(T).

Here, a indicates how much the amplitude increases by one pitch. In the filter 110, a gain controller 113 is provided at the output of the delay circuit 111, and when the output of the delay circuit 111 is multiplied by a, the subtraction circuit 1
Signals of approximately equal amplitude appear at the two inputs of 12. Therefore, even when the amplitude of the received signal, that is, the amplitude of the echo signal, is increasing, the filter 110 can sufficiently block the echo signal.

The above description is for the case where the signal amplitude on the receiving side is increasing, but when the signal amplitude on the receiving side is decreasing, the autocorrelation coefficient R(T) for one pitch apart is R(O).
becomes larger, and a=R(O)/R(T) becomes smaller than 1.

Therefore, the output of the gain controller 113 is attenuated, and the signal amplitudes of the two inputs of the subtraction circuit 114 are approximately equal, so that echo signals are sufficiently blocked.

Although only one stage of delay circuits is employed in the above embodiments, more effective echo removal performance can be expected by using a plurality of delay circuits. Furthermore, regarding the pitch extraction method, there are various other methods suitable for removing echoes.

Furthermore, when there is no received signal, the quality can be further improved by flattening the characteristics of the filter 110. In addition, conventional echo suppressor technology can be used as needed.

As described above, according to the present invention, a high-quality echo suppressor with a small device size and no cutting can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a first embodiment of the present invention, and FIG. 2 shows a filter 110 of the first embodiment.
FIG. 3 is an example of the pitch extraction circuit 120 of the first embodiment, FIG. 4 is a waveform diagram showing an example of determining autocorrelation, and FIG. 5 is an example of the second embodiment of the present invention. FIG. In the figure, 1 is a receiving side input terminal, 2 is a receiving side output terminal, 3 is a transmitting side input terminal, 4 is a transmitting side output terminal, 10 is an echo suppressor, 20 is a hybrid, 30 is a subscriber, 110 is a filter, 111
is a delay circuit, 112 is a subtraction circuit, 113 is a gain controller, 120 is a pitch extraction circuit, 121
1 is an AD converter, 122 is an autocorrelation circuit, 123 is a maximum value circuit, and 124 is a division circuit.

Claims (1)

[Scope of Claims] 1. An echo presser provided on the four-wire side of a two-wire/four-wire converter in a telephone line, which includes: (a) an analysis circuit for extracting the period T of the receiving side signal of the four-wire line; (b) four-wire line converter; A filter comprising: a delay circuit to which a transmission side signal of a line is supplied and whose delay amount is set to the T; and a subtraction circuit that outputs a difference signal between the transmission side signal and the output of the delay circuit; an echo suppressor for transmitting the filter output to a far end subscriber;