JPS59104700A - Voice starting apparatus - 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 The present invention provides sound/! 1. This invention relates to a voice cutting device that is used by cutting it into a response device or the like.

As shown in FIG. 1, a conventional voice response device basically consists of a voice recognition device 1 and an operating device 2 that operates based on the voice recognition result. voice recognition'! A station I consists of a voice collecting section M made of a microbon and a voice recognition section R. Here, the sound collected by the sound collecting JIS section M includes not only the sound S from the original sound source 3, but also so-called sound noise N from the surroundings. There is a risk that the collected noise N will be misrecognized by the 7"1 voice R1Z recognition unit and the operation mA will malfunction.For this reason, conventionally, it was recognized correctly] It was necessary to make various improvements to the recognition algorithm, such as using trilobal words that would otherwise be recognized, and taking into account commonalities between words that should be recognized.

The present invention has been made in order to solve the above-mentioned drawbacks of the prior art, and detects whether or not the voice generation source 3 is generating voice by its vibration, thereby inputting a signal to the voice recognition unit R. The purpose of the present invention is to provide an audio cutting device which can cut out an audio signal by cutting out the audio signal and prevent malfunctions caused by noise N.

Hereinafter, the present invention will be explained in detail with reference to the embodiment shown in FIG.

In this embodiment, a vibration detector I], a control circuit 4, and sections G1 to G are added to the conventional apparatus shown in FIG. The vibration detector is directly bonded to a part (for example, a vibrating part) of the sound generation source 3 consisting of a speaker. A storage gauge or the like can be used as the vibration detector.

The output waveform of the vibration detection section is, for example, as shown in FIG. That is, the vibration state is determined by the determining section Cd, but two problems occur at this time. The first problem is
Although it is originally in a vibrating state, since the signal level is low, it is judged as a non-vibrating state (mark A), so a state that should be called a pseudo non-vibrating state occurs. As a result, I accidentally changed the game i/part G to If7i (OF F)
This is the case. The second problem is due to the delay time of sound propagation due to the distance Nt between the sound generation source 3 and the sound collection device M, which causes the part of the goo 1 G to be cut off (OFI'') at an early stage.

In either case, the correct signal does not enter the recognition section R, causing malfunction. In order to prevent these, first of all, the vibration state determination section 0d of the gate control circuit 4 is connected to the vibration detector 1.
The output waveform (waveform in FIG. 3(a)) is binarized at a predetermined threshold value to obtain a binarized pulse (waveform in FIG. 3(b)). Next, the pseudo non-vibration state erasing unit 0e generates a pulse by delaying the falling part of the binarized pulse obtained by the vibration state determination unit 0d by a predetermined width W, as shown in FIG. 3(c). obtain.

Such pulses can be generated, for example, by using a monostable multivibrator.

In addition, by detecting the pulse width of the binarized pulse below a certain level and removing it from the binarized pulse,
The pseudo vibration state may be removed.

In addition, if the electrical signal of the audio collecting section M lags behind the output of the gate control circuit 4, a delay element may be provided between the vibration detector and the gate 1 section G to compensate for the delay. good. In the opposite case, a delay element may be provided between the gate section G and the audio collection section M to compensate for the delay. Further, the sound generation source 3 may be a speaker or a human. In the case of humans, the vibration detector is preferably placed on the person's throat or cheek.

As described above, the present invention allows audio to be extracted even when the level ratio (S/N ratio) between the audio S and the sound noise N is relatively small.

[Brief explanation of the drawing]

Claims (1)

[Claims] a transduction means for converting the sound from the sound generation source into an electrical signal; a detection means for detecting vibrations of the sound generation source;"-7
1. A voice cutting device, which is coupled to a converting means and a detecting means, and includes means for cutting out the output of the converting means based on the output of the detecting means.