JPS6316721A - Noise eliminating device - Google Patents

Abstract

PURPOSE:To eliminate the artificiality due to noise elimination by providing an adaptive difference predicting device and a means switching a difference signal of the adaptive difference predicting device or an input signal into a pseudo signal according to an instruction signal for a noise period so as to eliminate the noise and to interpolate the noise period by a simple constitution. CONSTITUTION:The adaptive difference predicting device 1 and the means 2 switching an input signal or the difference signal into a pseudo signal are provided. The adaptive difference predicting device 1 consists of an adder 5 connected to an input terminal 3, an adder 6 connected to an output terminal 4, an adder 7 outputting a predictive value, a transfer function section(Hz)8 for zeros, a transfer function section (Hp)9 for poles, and a predictive coefficient revision circuit 10, and the predictive coefficient revision circuit 10 controls the transfer function sections 8, 9. The means 2 consists of a pseudo signal output section 11 and a switch section 12, a pseudo signal from the pseudo signal output section 11 is applied as the difference signal or the input signal at the noise period via the switch section 12 and a signal being the result of interpolating the noise period is outputted from an output terminal 4.

Description

[Detailed Description of the Invention] (Summary) The input signal or difference signal of the adaptive difference predictor is switched to a pseudo signal in the noise section of the input signal, and the noise section is interpolated and output. It is possible to remove noise and, in place of the noise, output a signal that approximates the input signal immediately before the noise section.

[Industrial application field]

The present invention relates to a noise removal device that removes and interpolates a noise section when an input signal includes a noise section.

In wireless communication systems, the signal-to-noise ratio often deteriorates due to multipath fading, and by detecting the noise section where the signal-to-noise ratio has deteriorated and removing that noise section, it is possible to improve the voice signal. It is adopted to prevent quality deterioration. In this case, it is desired not only to simply remove the noise section, but also to perform interpolation so as not to give any unnaturalness to the audio signals before and after the noise section.

(Prior Art) A receiving device in a wireless communication system includes, for example, as shown in FIG. Section 73 and detection.

It is composed of a low frequency section 74 that drives a speaker by amplification etc. to reproduce the received audio signal, and a noise section detection section 75 that detects deterioration of the received signal due to multipath fading etc. and determines it to be a noise section. This noise section detection section 75
A switch that forcibly blocks the received signal is controlled by
A notch section 76 is provided in the low frequency section 42. Therefore,
A noise section where the received signal is abnormally deteriorated is forcibly cut off, and the noise is not reproduced and output from the low frequency section 74.

In this way, if only the noise section is blocked, the audio signal will be interrupted momentarily, and the quality of the audio reproduced and output from the speaker etc. will deteriorate.

Therefore, in order to interpolate the noise section, a configuration is known in which the signal in the noise section is passed through a low-pass filter.

In this case, the cutoff frequency of the low-pass filter becomes a problem.

Furthermore, a configuration has been proposed in which the audio signal immediately before the noise section is accumulated and the accumulated signal is repeatedly output during the noise section.

[Problems to be Solved by the Invention] As described above, the configuration that simply blocks out the noise section has the drawback that unnaturalness occurs in the reproduced audio signal. On the other hand, by interpolating the noise section by some means, the unnaturalness of the reproduced audio signal can be alleviated. In this case, the configuration using the conventional low-pass filter described above has the disadvantage that the quality of the reproduced audio signal deteriorates because the high-frequency components of the audio signal are suppressed.

Further, in a configuration in which the audio signal immediately before the noise section is stored, the memory capacity becomes large and the configuration and control become complicated because the audio signal is constantly written.

The present invention removes noise with a relatively simple configuration, and
The purpose of this is to interpolate the noise section and eliminate the unnaturalness caused by noise removal.

[Means for solving problems]

The noise removal device of the present invention uses an adaptive difference predictor, and will be explained with reference to FIG.

It has an adaptive difference predictor 1 and a means 2 for switching an input signal or a difference signal to a pseudo signal, and the adaptive difference predictor 1 includes an adder 5 connected to an input terminal 3 and an adder 5 connected to an output terminal 4. adder 6 and adder 7 that outputs the predicted value.
, zero transfer function section (Hz) 8, pole side transfer function section (Hp) 9, and prediction coefficient update circuit IO,
The transfer function unit 8.9 is controlled by the prediction coefficient update circuit 10. Further, the means 2 includes a pseudo signal output section 11 and a switch section 12, and in the noise section, the pseudo signal output section 1
The pseudo signal from 1 is added as a difference signal or an input signal via the switch section 12, and a signal obtained by interpolating the noise section is output from the output terminal 4.

[Effect]

During normal operation, the input signal from the input terminal 3 and the prediction signal from the adder 7 are added to the adder 5, a signal of their difference is output, and the difference signal and the prediction signal are added to the adder 6. The signals are added together and an output signal corresponding to the input signal is derived from the output terminal 4. Further, the prediction coefficient update circuit 10 is operated by the difference signal, and the transfer function units 8 and 9 are controlled by the prediction coefficient update circuit 10, so that equalization of the input signal is performed.

In the noise section, the input signal ' from the input terminal 3 is disconnected, and the pseudo signals such as the pseudo difference signal, impulse signal, white noise signal, and storage difference signal from the pseudo signal output section 11 are output from the switch section 12. or a pseudo input signal is applied to the input terminal 3 via the switch section 12, and a signal close to the input signal immediately before the noise section is output from the output terminal 4, and the signal is applied to the input signal immediately before the noise section. can be interpolated.

〔Example〕

FIG. 2 is a block diagram of an embodiment of the present invention, where 21 is a zero transfer function section (Hz), and 22 is a pole-side transfer function section (
23 is a prediction coefficient update circuit, 24 to 27 are adders, 28 and 29 are variable gain amplifiers, 30 is an input signal cutoff section, 31 is a control section that controls each section, and 32 is a pseudo difference signal generator. .

In this embodiment, the pseudo difference signal from the pseudo difference signal generator 32 is used as the difference signal for the adaptive difference predictor in the noise interval.

During normal operation, the input signal cutoff section 30 is controlled to be on by the control section 3, and the input signal a is applied to the adder 26. At this time, the control section 31 controls the gain of the variable gain amplifier 28 to 1 and the gain of the variable gain amplifier 29 to 0. Therefore, the input signal a and the predicted signal S are added to the adder 26, and the difference signal C as a result of their subtraction is output from the adder 26. This difference signal C is also applied to the variable gain amplifier 28 and the adder 27. Through the transfer function section 21. It is added to the prediction coefficient update circuit 23 and adder 25.

The adder 25 adds the difference signal C and the predicted signal S, and sends out an output signal d corresponding to the input signal a. This output signal d is applied to the transfer function section 22. Transfer function part 2
The output signals of 1.22 are added by an adder 24 to form a predicted signal S.

For example, the noise section is detected by the noise section detection section 7 in FIG.
5, and its detection signal f is applied to the control section 31. Based on this detection signal f, the control section 31 controls the input signal cutoff section 30 to be in an OFF state, and the variable gain amplifier 28
The gain of the variable gain amplifier 29 is sequentially controlled from 1 to 0, and conversely, the gain of the variable gain amplifier 29 is sequentially controlled from 0 to 1. Thereby,
Difference signal C from adder 26 and pseudo difference signal generator 32
The pseudo difference signal e from is gradually switched. On the other hand, when transitioning from a noise section to a normal signal section, the gain of the variable gain amplifier 28 is sequentially controlled from O to 1, and the gain of the variable gain amplifier 29 is sequentially controlled from 1 to O, so that there is no abrupt switching. do it like this. In the noise section, the control unit 31 controls the prediction coefficient update circuit 23 to stop its operation or to slow down the update speed.

Therefore, in the noise interval, a pseudo difference signal that approximates the state immediately before the input signal is cut off is added as the difference signal of the adaptive difference predictor, so that the output signal d corresponding to the input signal immediately before the noise interval is added as the difference signal of the adaptive difference predictor. will be output from the adder 25.

FIG. 3 is a block diagram of the adaptive difference predictor, and shows a case where the adaptive difference predictor has a sixth-order zero point and a third-order pole. 41
is zero transfer function part, 42 is pole side transfer function part, 43~
45 is an adder, 46 is a prediction coefficient update circuit, 47.48 is an adder, 49 to 57 are delay elements (Z-'), 58 to 6
6 is a coefficient multiplier, which adds the input signal X and the predicted signal X° to an adder 43, and outputs a difference signal E (n) of (x-X')=E(n). This difference signal E (n) is
The signals are sequentially delayed by delay elements 49 to 54 of transfer function unit 41 and applied to coefficient multipliers 58 to 63, and the respective delayed signals are applied to prediction coefficient update circuit 46. Further, the difference signal E (nl) from the adder 43 is applied to the prediction coefficient update circuit 46 and the adder 44.

By this adder 44, the difference signal E fn) and the predicted signal
′ is performed, and E(n) + x ′ = 5(n
) output signal is obtained. This output signal S (n) is
The signals are sequentially delayed by the delay elements 55 to 57 of the transfer function section 42 and applied to the coefficient multipliers 64 to 66, as well as to the prediction coefficient update circuit 46.

The prediction coefficient update circuit 46 includes a coefficient unit 58 of the transfer function unit 41.
63 and the coefficient sections 64 to 66 of the transfer function section 42, respectively, and the difference signal E (n)
The coefficient multipliers 58 to 63 are controlled by determining the correlation between the output signal S(n) and the difference signal E(n) by the delay elements 55 to 57. ), and the coefficient unit 64~
66 respectively.

The output signals of the coefficient units 58 to 63 are added by the adder 47, and the output signals of the coefficient units 64 to 66 are added by the adder 48, and the respective addition output signals are added to the adder 45, and the addition output is This becomes the predicted signal X'.

In the adaptive difference predictor, when the input signal X is completely equalized, the difference E between the input signal X and the predicted signal
(nl has no redundancy and becomes uncorrelated. That is,
It becomes white noise. Therefore, in a state where the input signal It can be output from the device 44.

Taking advantage of this point, it is also possible to replace the pseudo difference signal generator 32 in FIG. 2 with a white noise generator and interpolate the noise interval.

FIG. 4 is a block diagram of another embodiment of the present invention;
The same reference numerals as in the figure indicate the same parts, 33 is a pitch period detector, and 34 is an impulse generator. In this embodiment, the pitch period detector 33 detects the pinch period of the input signal a, and the impulse generator 3 detects the pinch period of the input signal a at a period corresponding to the pitch period.
4 to generate an impulse g. and,
In the noise interval, the impulse g from the impulse generator 34 is added via the variable gain amplifier 29 via the adder 27 instead of the difference signal from the adaptive difference predictor.

When the adaptive difference predictor is equalized to the input signal, the difference signal becomes white noise as described above, but by using the impulse g corresponding to the pitch period of the speech signal as the difference signal, it is possible to detect the speech just before the noise section. Output signal d that approximates the signal
It becomes possible to form a noise interval, and it is possible to interpolate the noise interval so that it becomes a natural state.

FIG. 5 is a block diagram of still another embodiment of the present invention,
35 is a differential signal accumulating unit, and 36 is an amplifier, and the same reference numerals as in the other figures in FIG. On the second day, a predetermined capacity is sequentially stored in the differential signal storage section 35 via the adder 27 and the amplifier 36. In the noise section,
The input signal a is cut off, and the differential signal stored in the differential signal storage section 35 is read out. Also, variable gain amplifier 29
The gain of C is controlled from O to 1, contrary to the gain of the variable gain amplifier 28, and the accumulated difference signal C is added to the adder 27, replacing the difference signal C that had been added via the variable gain amplifier 28. is added to the adder 25 and the transfer function section 21. Therefore, an output signal d corresponding to the input signal immediately before the noise period is obtained from the adder 25.

FIG. 6 is a block diagram of still another embodiment of the present invention,
37 is a pseudo original signal generator, and the same reference numerals as in the other parts of FIG. 2 indicate the same parts. By equalizing the adaptive difference predictor with the human input signal a, the adaptive difference predictor itself retains the frequency components of the input signal a, and therefore blocks the input signal a in the noise interval and generates a pseudo A pseudo signal i from the original signal generator 37 is added via the adder 27 instead of the input signal a. Even if this pseudo signal i is not sufficiently approximated to the input signal a, by passing the pseudo signal i through the adaptive difference predictor, the output signal d, which is the frequency component of the original human input signal a, is sent to the adder. It will be output from 25.

It is also possible to have a configuration in which the output signal d is accumulated as the pseudo original signal generator 37, and the accumulated output signal d is added instead of the input signal a during the noise interval.

Furthermore, by controlling the gains of the variable gain amplifiers 28 and 29 by the control section 31, it is possible to prevent abrupt switching between the noise section and the normal section.

〔Effect of the invention〕

As explained above, the present invention uses the adaptive difference predictor 1, blocks the input signal in the noise interval, and adds a pseudo signal via the means 2 instead of the difference signal of the adaptive difference predictor or the human input signal. When the adaptive difference predictor is equalized to the input signal, it retains the frequency components of that input signal, and when a pseudo signal is added as an input signal or a difference signal, the components of the input signal Since it is possible to obtain an output signal having , there is an advantage that noise due to multipath fading etc. can be removed and interpolation can be performed so that the noise section is continuous in a natural state.

[Brief explanation of drawings]

Figure 1 is a block diagram of the principle of the present invention, Figure 2 is a block diagram of an embodiment of the present invention, Figure 3 is a block diagram of an adaptive difference predictor, and Figures 4 to 6 are different versions of the present invention. FIG. 7 is a block diagram of the main part of the receiving device. 1 is an adaptive difference predictor; 2 is means for outputting a pseudo signal; 3
is an input terminal, 4 is an output terminal, 5 to 7 are adders, 8 is a zero transfer function section, 9 is a pole side transfer function section, 10 is a prediction coefficient update circuit, 11 is a pseudo signal output section, and 12 is a switch part,
21 is a tower side transfer function section (Hz), 22 is a pole side transfer function section (Hp), 23 is a prediction coefficient update circuit, 24 to 27 are adders, 28.29 is a variable gain amplifier, 30 is an input signal 31 is a control section that controls each section, and 32 is a pseudo difference signal generator.

Claims (1)

[Claims] A noise removal device for interpolating a noise section of an input signal, comprising: an adaptive difference predictor (1);
1. A noise removal device comprising: means (2) for switching the differential signal or input signal of 1) to a pseudo signal.