JPH0786760B2 - Voice recognition system - Google Patents

Description

TECHNICAL FIELD The present invention relates to a speech recognition system.

[Conventional technology]

Conventionally, a voice recognition system is configured, for example, as shown in FIG. That is, the voice recognition system 1 appropriately processes the microphone 3 that picks up the voice of the speaker 2 and converts it into an electric signal, the amplifier 4 that amplifies the signal from the microphone 3, and the output signal from the amplifier 4. And a signal processing circuit 5 for performing voice recognition.

According to the voice recognition system 1 configured as described above, the voice of the speaker 2 is input from the microphone 3 to the signal processing circuit 5 via the amplifier 4, and the signal processing circuit 5 performs predetermined voice recognition processing. Is performed, the processing result is output.

[Problems to be Solved by the Invention] However, in the voice recognition system 1 having such a configuration, even if a microphone having high directivity is used as the microphone 3, the voice of the speaker 2 and the surroundings are not used. Since foreign sounds such as noises in the microphone also enter the microphone 3, there is a problem in that the recognition rate is lowered due to the mixing of the foreign sounds, which makes voice recognition difficult and erroneous recognition occurs. was there.

In view of the following points, it is an object of the present invention to provide a voice recognition system that improves the voice recognition rate by eliminating the influence of external sound as much as possible.

[Means for Solving the Problems]

According to the present invention, the above-described object is to amplify the output signals of a plurality of microphones arranged such that the directional characteristics of the speakers overlap each other at the speaker's position and the voices of the speakers are input equally. The absolute value component of the difference between the output signals from each rectifier circuit is subtracted from the sum of the amplifiers for rectifying, the rectifier circuit that half-wave rectifies the output signals from each amplifier, and the output signal from each rectifier circuit. It is achieved by including a pre-processing circuit for performing a voice recognition process and a signal processing circuit for performing a voice recognition process based on a signal from the pre-processing circuit.

[Work]

According to the present invention, the voice of the speaker is picked up by a plurality of microphones, the components having the same amplitude and phase among the output signals from the respective microphones, that is, the signals relating to the voice of the speaker are added, and the amplitude and the phase are matched. By subtracting a signal relating to a component that does not exist, that is, an external sound, etc., a signal to be a target of voice recognition is emphasized, and a signal such as an external sound that interferes with the voice recognition is offset and attenuated. Therefore, the S /
Since the N ratio is improved, the voice recognition process can be surely performed, which improves the voice recognition rate.

〔Example〕

Hereinafter, the present invention will be described in more detail based on the embodiments shown in the drawings.

FIG. 1 shows a speech recognition system according to the present invention,
The voice recognition system 10 amplifies the output signals of a plurality of microphones 12a and 12b (two in the case of illustration) and the microphones 12a and 12b, which are arranged so that the directional characteristics of the speakers 11 overlap each other. Amplifiers 13a, 13b for, and the output signal from each amplifier 13a, 13b, the pre-processing circuit 14 that adds for components whose amplitude and phase match and attenuates and removes components whose amplitude and phase do not match, The signal processing circuit 15 is configured to perform a voice recognition process based on the signal from the preprocessing circuit 14.

Specifically, the preprocessing circuit 14 half-wave rectifies the signals A and B (see FIG. 2B) from the amplifiers 13a and 13b as shown in FIG. , B'rectifier circuit 21a,
21b, an adder circuit 22 for adding the signals A ', B', a subtractor circuit 23 for subtracting the signals A ', B', and a rectifier circuit 24 for rectifying an absolute value of the output signal of the subtractor circuit 23, The adder circuit 22
Of the rectifier circuit 24 and the output signal of the rectifier circuit 24.

The embodiment of the present invention is configured as described above, and the microphone 12
The audio signals picked up by a and 12b are respectively amplified by amplifiers 13a and 13b, and then input to the preprocessing circuit 14, and in this preprocessing circuit 14, the signals A and B from the respective amplifiers 13a and 13b are Half-wave rectified by the rectifier circuits 21a and 21b respectively, and the signals A ',
B ', which is added by the adder circuit 22 to obtain the signal (A'
+ B '). Further, the subtraction circuit 23 subtracts (B'-A '), and the absolute value rectification by the rectification circuit 24 becomes | B'-A' |. Then, the subtraction circuit 25 subtracts it to obtain a signal. {(A '+ B')-| B '
−A ′ |}.

Here, the voice of the speaker 11 is captured by the both microphones 12a and 12b as signals of the same amplitude and the same phase, and the
13a, 13b are equally amplified, and then input to the preprocessing circuit 14, but because the amplitude and phase match,
In the adder circuit 22, the signals are added to each other to give an emphasized signal, while in the subtractor circuit 23, they are canceled.

In addition, external noise such as noise can be heard by both microphones 12a and 12b.
Is regarded as a signal whose amplitude and phase do not match, and is therefore added to each other in the adder circuit 22 to be a emphasized signal, but also in the subtractor circuit 23, it is added to the − side, and after absolute value rectification, By subtracting from the signal from the adder circuit 22, the signal is attenuated.

At this time, the external sound picked up by only one of the microphones 12a or 12b is completely canceled and disappears, as shown by the signal waveform at the left end of FIG. 2 (B). In the case of signals having different levels in 21b, they can be significantly reduced by cancellation, and the S / N ratio of the output signal from the pre-processing circuit 14 is increased, so that the voice recognition processing by the signal processing circuit 15 can be reliably performed. Can be done.

〔The invention's effect〕

As described above, according to the present invention, the voice of the speaker is picked up by a plurality of microphones, and the components having the same amplitude and the same phase among the output signals from the respective microphones, that is, the signals relating to the voice of the speaker are added. In addition, by subtracting a signal related to a component whose amplitude and phase do not match, that is, an external sound, etc., the signal that should be the target of the voice recognition is emphasized and the signal such as the external sound that interferes with the voice recognition is canceled. Since the signal is attenuated, the S / N ratio of the signal subjected to the voice recognition process in the signal processing circuit is improved, and the voice recognition process can be surely performed. Therefore, the voice recognition rate is improved.

Thus, according to the present invention, it is possible to provide an extremely excellent voice recognition system in which the voice recognition rate can be improved by eliminating the influence of extraneous sound as much as possible.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram showing an embodiment of a voice recognition system according to the present invention, FIG. 2 shows a concrete configuration example of a preprocessing circuit in the embodiment of FIG. (A) is a block diagram, (B)
3 is a graph showing an example of a voice waveform in each block. FIG. 3 is a schematic diagram showing an example of a conventional voice recognition system. 10 …… Speech recognition system; 11 …… Speaker; 12a, 12b …… Microphone; 13a, 13b …… Amplifier; 14 …… Preprocessing circuit; 15 …… Signal processing circuit.

Claims (1)

[Claims] 1. A plurality of microphones arranged so that the directional characteristics of the speakers overlap with each other and the voices of the speakers are equally input, and an amplifier for amplifying an output signal of each microphone. , A rectifier circuit that half-wave rectifies the output signal from each amplifier, and a preprocessing circuit that subtracts the absolute value component of the difference between the output signals from each rectifier circuit from the sum of the output signals from each rectifier circuit And a signal processing circuit that performs a voice recognition process based on a signal from the preprocessing circuit.