JPS62230229A - Echo eliminator - Google Patents

Abstract

PURPOSE:To eliminate noise produced around the boundary of each band by dividing the band into plural numbers, converting each band into a low frequency, eliminating an echo signal, restoring the frequency into the original band so as to eliminate the echo signal of all bands. CONSTITUTION:When a transmission/reception discrimination circuit 9 discriminates the mode as the reception mode, the circuit 9 throws a switch 11 to the position B and throws a switch 13 to the position D. As a result, a reception signal inputted from a terminl 23 is outputted from a speaker 17 via a band stop filter 5, a band split echo canceller 3 and an amplifier 15. In this case, the band stop filter 5 eliminates the frequency component of the band near the boundary of the band to be split by the band split echo canceller 3, then the noise around it is eliminated by using a quadrature mirror filter. The noise is elimlnated almost similarly in case of the transmiaaion.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Industrial Application Field) The present invention is directed to a loudspeaker telephone, in which the sound emitted from the speaker is reflected by the input wall and enters the microphone, causing acoustic coupling. The present invention relates to an echo removal device that prevents howling from occurring.

(Prior art) Amplified telephones, which allow you to talk without using a handset with the sound amplified from a speaker, are very convenient devices that allow you to use both hands freely during a call, and are especially useful for conference calls, which have been attracting attention in recent years. is considered to be indispensable.

In this type of public address telephone, the incoming sound is amplified by an amplifier and sent to the speaker, so the sound emitted from the speaker is picked up by the microphone, amplified by the equipment amplifier, and this signal is transmitted back to the receiver with the sidetone etc. generated by the hybrid coil. A signal loop occurs due to the input to the amplifier. Therefore, simply the microphone,
If a loudspeaker telephone is configured with a speaker and an amplifier, this signal loop may cause howling, making it impossible to make a call.

Conventionally, in order to prevent this howling, a voice switch system has been widely used for loudspeaker telephones to break this loop by inserting a loss in the receiving side when transmitting a voice and in the transmitting side when receiving a voice. However, this system has the disadvantage that simultaneous communication is not possible due to the loss introduced into one of the communication paths, and that it gives the speaker an unnatural switch feeling when the loss is switched.

On the other hand, with advances in digital signal technology, echo cancellers have become relatively easy to implement using LSI and the like, and have attracted attention as a technology that can replace the voice switch method.

This echo canceller 300 is used in a loudspeaker telephone as shown in FIG.
(t) +y (t), in order to cancel only the signal y(t) that comes out from the speaker 302 and is reflected by the wall etc.,
The signal loop mentioned above is broken between the speaker and the microphone, and howling can be prevented. Therefore, if this echo canceller 300 is used, there is no need to introduce loss into the communication path, so better speech quality can be obtained compared to the voice switch system.

This echo canceller can be configured as shown in FIG. 4, for example.

In the figure, the transversal filter 401 converts the signal X(t) from the speaker into y(t) due to reflection.
(this is called a transfer function of the echo path) and a filter with approximate characteristics (this filter is called a pseudo-echo path), and is generally a transversal filter that has tap coefficients that approximate the impulse response of the transfer function. Consists of filters. In this circuit, due to the input @y(t),
A pseudo echo signal y(t) which is an approximation of the reflected signal @y(t) is generated, and from the microphone input signal y(t)+n(t), the subtractor 402 calculates V(t)(:y(t) ) by subtracting only the echo signal y(t) and the true transmitted signal n
Extract (t).

Here, the tap coefficients of the transversal filter 401 are determined sequentially by a widely known learning algorithm such as the learning identification method, using the signal x(t) from the speaker and the output signal e(t) of the bond reducer 402. is required. In the figure, an estimation circuit 403 inputs the tap coefficients from the filter 401, modifies them according to the learning identification method, and returns them to the filter 401 again. By sequentially performing this process, the tap coefficients of the transversal filter are finally The coefficients approximate the impulse response of the transfer function of the reflected signal.

Note that this tap coefficient estimation needs to be performed only when the microphone input is the reflected signal y(t). Otherwise, the transmitted signal n (t) will mask the reflected signal y (t),
Tap coefficient estimation becomes inaccurate.

Therefore, the double talk detector 404 compares the power of the microphone input signal n (↑) + y (t>) and the receiver signal X (1), and the power of n (t) + y (t) is x (
When the power of t) is greater than the value reduced by a certain value, it is determined that the transmission signal n (t) is present, and updating of the tap coefficient of the estimation circuit 403 is prohibited.

In order to use the above-mentioned echo canceller to remove acoustic coupling in a public address telephone, if it is configured as a transversal type with a long acoustic echo impulse response length of 300 m5 to 400 m5, a sample of about 1B-kHz will require 3000 to 4000 m5. A tap or so is required, leading to an increase in the amount of hardware.

For this reason, it is common to construct such an adaptive transversal filter with a long tap length by dividing it into multiple echo cancellers. Configure using multiple pieces.

On the other hand, in a high-quality loudspeaker telephone, the sample frequency is 16
kHz is common, and in this case, the number of performance steps in one sample period is halved, so the tap length that can be achieved with one DSP chip is also halved, and the delay processing per tap is DS because the time will be halved
The length of the echo that can be processed per 1 dip of P becomes 1/4 when the band is doubled (the sampling frequency is doubled).

-〇- Therefore, with the same echo processing capacity, four DSPs in the 3.4kHz band are required in the 7k) Iz band. For this reason, a band-splitting type echo canceller as shown in Fig. 5 has been proposed [Practical Study of Space Propagation Type Echo Cancellation Device, Institute of Electronics and Communication Engineers, Information Systems Division National Conference No.
247, for NTT Yoshikawa et al.

This method uses, for example, a received signal x(t) in the 0 to 8 kHz band.
After being digitized by an A/D converter 5 echo canceller, a low pass filter 503 and a bypass filter 50
5 into a low band of 0 to 4 kHz and a high band of 4 to 8 kHz, and each sampling section 507 and 509 generates 8 kt.
(After sampling at z, it is input to an 8kHz echo canceller 511, 513.
5, a sampling section 51 sends a signal of 8 kHz to each
7.519 to a signal with 16 kHz interpolated samples (
In other words, 0 is interpolated alternately) and the low-pass filter 521
, a bypass filter 523, an adder 525 adds these outputs, a D/A converter 527 converts the output into an analog signal, and the output is output from a speaker 517.

A signal containing an echo y (t>) reflected from a wall etc. is picked up by a microphone 529 and digitized by an A/D converter 531.
Band-divided by bypass filter 533 and low-pass filter 535, and 8kHz by sampling section 537.539
After being sampled in the echo canceling subtractor 541.5
43 subtracts the pseudo echo signal output from the echo cancellers 511 and 513 to remove the echo component, and then the sampling section 545 and 547 output the signal at 16 kHz.
D
/A converter 555 converts it into an analog signal and outputs it.

As described above, in this method, the signal is bypass filtered as shown in Figure 6a on the high band side, for example, at 4 to 8 kHz.
After that, as shown in the figure, the signal becomes 4kHz.
z, and can be folded back to the low frequency range 0 to 4 kHz. After processing the echo canceller in this low frequency range, as shown in Figure d, the processing result O to 4kHz sampling fold is extracted as a component of 4 to 8kHz using a bypass filter (z component), thereby converting the high frequency signal. After processing the signal in the low band, it can be returned to the high band.

(Problem to be Solved by the Invention) However, in such a method, if the low-pass filter and bypass filter for band division do not sufficiently cancel out signals outside a predetermined band, out-of-band components will be generated when folding back by sampling is performed. There is a problem in that signal distortion occurs because the signal is folded back within the required band.

For this reason, it is necessary to use filters with characteristics that cause a large loss of about 80 dB outside the band as shown in Figure 7 for such low-pass filters and bypass filters, but it is difficult to use filters with such characteristics in real hardware. In order to realize this, it is necessary to have portions with changing characteristics such as Δδ■ and Δδ[ shown in FIG.

Therefore, a gap of Δδ[10 to δH−100 to 200 Hz] is generated in the signals passed through these band division filters, which becomes a major factor in deterioration of speech quality.

To prevent this, as shown in Fig. 8, a band division filter is used, in which the phase characteristics and amplitude characteristics of the low-pass filter and the bypass filter are symmetrical.
r: A Filter Family Design
ed for Use Quadrature Mir
ror Fi 1ter Bank: 1980 IEE
E ICASSP P271 to P273) are used.

``This filter has out-of-band components, but when the out-of-band components of the low-pass filter and bypass filter are added, the result becomes zero, so the band gap as described above does not occur, and as a result, Flat characteristics can be obtained at 0 to 8 kHz.

As described above, quadrature mirror filters have very excellent characteristics, but when used together with echo cancellers, the following problems occur, which is a major obstacle to their introduction into band-splitting echo cancellers. It became.

That is, echo cancellers 511 .
513, the folding of out-of-band components by the quadrature mirror filter (shaded area in FIG. 8) becomes a noise component, resulting in a problem that the amount of cancellation is degraded as shown in FIG. 9.

For this reason, in loudspeaker telephones, howling and echoes occur near the boundaries of the band.

The present invention has been made to solve these conventional problems.

[Structure of the Invention (Means and Effects for Solving the Problem)] The present invention provides a device for removing echo signals that are re-inputted into the transmitting signal device due to echoes from the audio signal output from the receiving device. A band division type echo that divides the band into multiple parts using a quadrature mirror filter, converts each band to a low band, then removes the echo signal, returns to the original band, and removes the echo signal of all bands. - It is characterized by comprising a canceller and a plurality of band elimination filters that block frequencies near the boundaries of each band to be divided in the echo canceller.

(Function) When a quadrature mirror filter or the like is used as an echo canceller, noise generated near the boundaries of each band to be divided can be removed.

(Example) Embodiments of the present invention will be described below based on the drawings.

FIG. 1 is a block diagram of the construction of a removal device according to an embodiment of the present invention. In the figure, this echo canceling device 1 is comprised of a band division type echo canceller 3, a band rejection filter 5.7, a transmission determination circuit 9, and a switch 11.13. The above-mentioned quadrature mirror filter is used for the band division echo canceller 3, and it includes an amplifier 15, a speaker 17, an amplifier 19
, microphone 21 is connected.

The band division type echo canceller 3 is an echo canceller having a plurality of quadrature mirror filters as described above. Band-stop filter 5.7
is a filter that has characteristics as shown in FIG. 2, that is, removes frequencies near the boundary to be divided by the echo canceller 3 (approximately 4 kHz in this figure). As this band rejection filter 5.7, a widely known simultaneous Chebyshev type filter or the like is used.

The transmission/reception determination circuit 9 determines the levels of the transmission signal S and the reception signal R, and controls the switches 11 and 13.

Next, the operation of this embodiment will be explained.

The transmitting/receiving determination circuit 9 compares the signal levels of the transmitting signal S and the receiving signal R, and if it determines that the call is being received, connects the switch 11 to the B side and the switch 13 to the D side. As a result, the reception signal inputted from the terminal 23 is outputted from the speaker 17 via the band rejection filter 5, the band division type echo canceller 3, and the amplifier 15. Since the band-elimination filter 5 removes frequency components near the boundaries of the bands to be divided by the band-splitting echo canceller 3, it is possible to remove noise in the vicinity caused by using a quadrature mirror filter.

When the transmission/reception determining circuit 9 determines that it is time to transmit, the switch 11 is connected to the A side and the switch 13 is connected to the C side. As a result, the audio input from the microphone 21 is outputted from the terminal 25 via the amplifier 19, the band division echo canceller 3, and the band rejection filter 7. In this case as well, by passing the signal through the band rejection filter 7, noise components in bands near the boundaries of the bands to be divided by the quadrature mirror filter can be removed, as in the case of transmission.

Therefore, according to this embodiment, each transmitting and receiving signal is regenerated without a band gap by the band-splitting echo canceller using a quadrature mirror filter, and it is possible to prevent deterioration of the amount of cancellation near the band-splitting boundary. Therefore, you can make good calls.

Note that the band rejection filter 5.7 may have characteristics as shown by the dotted line in FIG. 2.

[Effects of the Invention] As described above in detail, according to the present invention, it is possible to remove noise occurring near the boundaries of each band to be divided in a band division type echo canceller.

[Brief explanation of drawings]

FIG. 1 is a block diagram of the configuration of an echo canceling device according to this embodiment, FIG. 2 is a characteristic diagram of a band rejection filter, and FIG. 3 is a configuration diagram of a public address telephone using a conventional echo canceller.
Figure 4 is a block diagram of an echo canceller, Figure 5 is a block diagram of a loudspeaker telephone using a band division echo canceller, Figure 6 is a characteristic diagram of the loudspeaker telephone shown in Figure 5, and Figure 7 is a general diagram. Figure 8 is a characteristic diagram of a band division filter using a quadrature mirror filter, and Figure 9 is a characteristic diagram of a band division echo canceller using a quadrature mirror filter. It is a cancellation characteristic diagram. 1...Echo removal device 3...
...Band division type echo canceller 5.
7...Band rejection filter frequency (KHz) Figure 2 (day ρ) im rise (
Japan p) Ministry # n meeting (Japan P) Iq town ear IQR-

Claims (2)

[Claims] (1) In a device that removes echo signals that are re-inputted into the transmitting signal device due to echoes from the audio signal output from the receiving device, the band is divided into multiple bands, each band is converted to a low frequency band, and then the echo signal is removed. Band-splitting echo signal that returns to the original band and removes echo signals in all bands.
An echo canceling device comprising: a canceller; and a plurality of band elimination filters that block frequencies near the boundaries of each band to be divided in the echo canceller. (2) The echo canceling device according to claim 1, wherein the band division in the band division type echo canceller is performed by a plurality of quadrature mirror filters.