JPH06101666B2 - Adaptive noise eliminator - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial application] The present invention is a noise elimination device for eliminating noise components from input signals from two input devices by using an adaptive filter and outputting a voice for detection. Regarding

[Conventional technology]

The adaptive filter used in this type of noise eliminator automatically changes the filter coefficient so as to minimize the root mean square of the difference between the input signal and other reference signals.

Therefore, when an adaptive filter is used as the noise removing device, the input device A installed near the sound source to be detected and far from the noise source and the input device B installed near the noise source and far from the sound source. From the two input devices consisting of
First, only noise is input, the coefficient of the adaptive filter is determined so as to estimate the noise signal in the input device A from the noise signal in the input device B, and then the noise component remains unchanged from the input device A. It is necessary to input the voice to be detected and to remove only the noise signal from the input signal using the filter coefficient determined previously.

[Problems to be solved by the invention]

In the conventional noise eliminator, when the speech to be detected is input without distinguishing between the step of determining the filter coefficient by inputting only the above noise and the step of performing noise removal by inputting the speech to be detected. Also updates the coefficient of the adaptive filter based on the input signal including the audio signal. For this reason, noise removal processing is performed by a filter coefficient that is unsuitable for removing only noise signals, and as a result, there is a drawback that the output voice is distorted.

In addition, it is necessary to determine whether or not an audio signal is generated by calculating the input level ratio from the noise source to the two input devices and then comparing it with the current input level ratio of the input device.

[Means for solving problems]

The adaptive noise elimination device of the present invention includes two detection devices that detect respective sound source levels of a voice signal source and a noise source,
An adaptive filter that inputs the output signals of the two detection devices and removes noise components, a switch for designating a section in which no voice signal is generated and setting a threshold value, and a state of the switch is the voice signal source. If there is no voice signal, the output ratio (threshold value) of the integrating means that integrates each of the output signals is calculated. If the output of each of the integrating means is input and the output of each of the integrating means is input, the output ratio is compared with the threshold value, and when the output ratio of the integrating means is larger than the threshold, it is determined as a voice section. Comparing means for stopping the adaptive operation of the filter coefficient of the adaptive filter.

〔Example〕

 Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention. Microphones 11A and 11B, low-pass filters 12A and 12B, a sample and hold circuit 13A, and a voice source A and a noise source B, which are detection targets,
When the input devices 1A and 1B composed of 13B and the analog / digital converters 14A and 14B are arranged as shown in the figure, the input levels of the sound source components closer to each other increase.

The threshold value setting switch 2 is for temporarily outputting "1" by an external means such as a speaker operation to set a threshold value when an audio signal is not generated from the audio source A,
"0" during normal voice input waiting state and voice signal generation
Holding After that, as long as there is no change in the position and characteristics of the noise source B or changes in the positions and characteristics of the microphones 11A and 11B, it is not necessary to reset the threshold value and "0" is held.

In the voice cutout unit 3, the input signals from the voice input device 1A and the noise source input device 1B are input to the integrators 31A and 31B. The outputs of the integrators 31A and 31B and the output of the threshold setting switch 2 are input to the threshold calculation circuit 32. If the output from the switch 2 is "1", the input signals from the input devices 1A and 1B are those of a noise source. Output ratio of integrators 31A and 31B,
That is, the average power level of the mixed noise detected by the voice source input device 1A and the average power level ratio of the noise detected by the noise source input device 1B are calculated and set as the threshold value for voice segment extraction, and the output from the switch 2 If the value is "0", there is a possibility that a voice signal is being generated because a voice input waiting state or a voice input is being performed. Therefore, the latest threshold value held is output without calculating the threshold value.

During normal use after the threshold is set, that is, when the output of the switch 2 is "0", the comparison circuit 33 cuts out the voice section of the signal detected by the voice source input device 1A. In the comparison circuit 33, the output ratio of the integrators 31A and 31B, that is, the ratio between the short-time average power level of the signal including the voice detected by the voice source input device 1A and that of the noise signal detected by the noise source input device 1B, And the threshold value for the voice segment cutout, which is the output of the threshold value calculation circuit 32, are compared, and if the former is smaller than or equal to the latter, it is determined that the detection target voice is not generated, and the comparison circuit
33 outputs "0".

On the other hand, the input signals from the input devices 1A and 1B are delayed by the delay circuits 4A and 4A.
Through B, the output of the comparison circuit 33 and the adaptive filter 5 at the same time
Entered in. In the adaptive filter 5, the filter coefficient is updated from the input signal when the output from the comparison circuit 33 of the audio cutout unit 3 is "0", and the filter coefficient is updated when the output from the comparison circuit 33 is "1". Of the input signal is canceled by using the latest filter coefficient and the output signal C
Is output.

〔The invention's effect〕

As described above, according to the present invention, in an environment in which noise and voice are mixed, input when only noise of two input devices is externally given by a signal switch input indicating a section in which no voice signal is generated By setting the level ratio as a threshold value, the presence or absence of speech, which is the detection target, is appropriately judged, and the updating operation of the coefficient of the adaptive filter is controlled accordingly, and the optimum filter coefficient is used to remove only noise. There is an effect that noise can be removed.

As a result, the distortion of the voice waveform that has passed through the adaptive filter for noise removal can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention. A ... voice source, B ... noise source, C ... noise eliminator output signal, 1A, 1B ... input device, 11A, 11B ... microphone, 12A, 12
B ... Low-pass filter, 13A, 13B ... Sample-and-hold circuit, 14A, 14B ... Analog / digital converter, 2 ... Threshold setting switch, 3 ... Voice cutout section,
31A, 31B ... Integrator, 32 ... Threshold value calculation circuit, 33 ... Comparison circuit, 4A, 4B ... Delay circuit, 5 ... Adaptive filter.

Claims (1)

[Claims] 1. A two detection device for detecting respective sound source levels of a voice signal source and a noise source, an adaptive filter for receiving output signals of the two detection devices and removing noise components, and a voice signal generation device. In the case where a switch for designating a non-existing section and setting a threshold value and the state of the switch indicate that there is no audio signal of the audio signal source,
Threshold value calculating means for calculating an output ratio (threshold value) of an integrating means for integrating each of the output signals, and holding a threshold value when there is no immediately preceding audio signal when the state of the switch is reversed; Comparing the output ratio of the integrating means and comparing the output ratio with the threshold value, and if the output ratio of the integrating means is larger than the threshold value, it is judged as a voice section and the adaptive operation of the filter coefficient of the adaptive filter is stopped. An adaptive noise elimination device comprising: