JPS607497A - Voice recognition equipment - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field The present invention relates to a speech recognition device that reduces the influence of noise.

<Prior Art> In conventional speech recognition, the input level of the microphone was manually set to an appropriate position by looking at a level meter, and the setting was reset if the number of incorrect responses increased due to noise. Those equipped with an automatic gain adjustment circuit are effective when used in low-noise locations, but pose problems when used in high-noise locations because they are easily affected by noise. As described above, manual gain adjustment is complicated, and the adjustment itself may be difficult. - Conventional automatic gain adjustment circuits are still effective when the noise is small and the voice power level fluctuates, but it is difficult to apply when the noise is large.

<Background of the Invention> There are many possible causes of how noise affects speech recognition and causes incorrect responses. Among these, the most common cause is an error in cutting out the input audio. Cutting out this input audio is an important function that determines the point at which input audio processing starts from the moment the input level exceeds a certain level value. Processing such as analysis is started from the middle of the audio signal and identification judgment is made.
An incorrect response or no response will result.

Therefore, in order to improve this input audio clipping error, it is necessary to set the audio clipping level to a value slightly higher than the noise level. , since the audio input signal becomes relatively small, there is a possibility of insufficient input, resulting in erroneous responses or no response. Therefore, in general, it is necessary to constantly increase the input voice, to bring the microphone closer to the mouth, or to speak loudly.

<Object of the Invention> The present invention provides means for automatically searching for and setting a value slightly higher than the noise level, and means for indicating and displaying an appropriate target value for the input audio level, in the above-described operations. By providing this, it is possible to reduce the influence of noise and obtain a stable response.

The most important point here is what is the difference between the noise level and the input audio level, what characteristics are used to determine the noise level, and how is the extraction level set based on that? 〉 An embodiment of the present invention will be described below with reference to the drawings.

In the figure, 1I''i microphone, 2id voltage-gain control microphone amplifier, 3 a noise level detection circuit,
4 is an input voice display device, and 5 is a voice recognition circuit.

First, after turning on the power switch or the microphone switch, voice input is usually not performed immediately, but often after a period of 2 seconds or more. During this time, noise is constantly being input from the microphone l. The noise level detection circuit 3 detects the noise level and holds it at a peak voltage, which is used as a control voltage for the voltage-gain control circuit of the microphone amplifier 2.

This control voltage is held at a peak value voltage, and is maintained at this value while there is input audio, but this value gradually decreases when there is no input audio. It would be more convenient if the rate of decline could be adjusted depending on the environment.

When the control voltage gradually decreases and the noise level exceeds the voltage control level again at a certain level, the peak is held at that noise level and the control pressure increases. However, if the noise level exceeds the voltage control level For example, if a value of Of course, the setting of the duration for determining the difference between this noise and the input voice is just an example, and different time settings are made depending on the type of noise, etc. If there is no input voice, the control voltage is set again. Gradually the voltage starts to drop and the voltage exceeding the control voltage reaches 0.
If the duration is 2 seconds or less, the noise level is held at its peak as a control voltage.

By repeating the above operations, a control voltage is obtained that automatically sets the optimum audio signal extraction level.

This control voltage is applied to the voltage-gain control circuit of the microphone amplifier 2 to automatically adjust the gain.

By the way, due to the above-mentioned cutout level setting, when the noise level is high, the gain of the microphone amplifier 2 is automatically reduced, and as a matter of course, the input audio level must be increased. Conversely, when the noise level is low, the gain of the microphone amplifier 2 becomes high, so that even relatively small voices can be recognized. However, even in the latter case, if the voice is too loud, the recognition rate may decrease.

The input audio display device 4 is intended to improve the two disadvantages caused by the automatic setting of the cutout level, and displays the appropriate input level range as well as the instantaneous input level.
To easily display how far you have deviated from your target level. If this is satisfied, the output of the microphone amplifier 2 is always input to the subsequent speech recognition circuit 5 at an appropriate level, so that a stable response can be obtained.

<Effects of the Invention> As described above, according to the present invention, there is no need to constantly increase the volume of human voice, and signals can always be input to the voice recognition circuit at an appropriate level as necessary, thereby reducing noise. A useful speech recognition device that reduces the influence and obtains stable responses can be provided.

[Brief explanation of the drawing]

The drawing is a block diagram showing an embodiment of the present invention. DESCRIPTION OF SYMBOLS 1... Microphone, 2... Microphone amplifier with voltage-gain control, 3... Noise level detection circuit, 4...
Input voice display device, 5... voice recognition circuit. Agent: Patent attorney Aihiko Fuku (2 others - ζ61)

Claims (1)

[Claims] 1. A microphone amplifier with voltage-gain control, means for detecting the noise level input from the microphone and controlling the gain of the microphone amplifier in order to set the cutout level of the audio input at any time, and according to the gain control. 1. A speech recognition device comprising means for simultaneously displaying a range of appropriate speech input levels and an instantaneous speech input level.