JPS60216400A - Voice section detecting circuit - 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 relates to a speech segment detection circuit in a speech recognition device, and more particularly to optimization of a speech segment extraction threshold.

[Examination] In general, when the target is input speech with a good signal-to-noise ratio, it is relatively easy to extract sections where speech exists. However, in an environment where a speech recognition device is actually used, speech is input in a form that contains various kinds of noise and is superimposed on the noise. Since the noise at this time changes from moment to moment, it is difficult to detect a stable voice section by setting a fixed threshold value and cutting out the voice section, which may cause misrecognition. When clipping is performed using such a fixed threshold, low-power parts such as the beginning and end of speech and silent consonants are cut. Furthermore, under high noise conditions, noise is added and cut out before and after what should normally be a voice section.

The present invention has been made in view of the above-mentioned circumstances.
In particular, it is an object of this invention to provide a speech section detection circuit that can perform stable speech section detection and ensure a stable recognition rate regardless of the level of ambient steady noise.

The configuration of the present invention will be described below based on embodiments.

A feature of the present invention is that the noise level at the time when a predetermined time t has elapsed from the end of the voice section is held, and this is set as the threshold f1 for detecting the sound-1 section. This time t is based on the fact that there is a silent interval of 200 to 4.00 ms, such as a word with a consonant, and if the next sound is input after this time, the preceding and succeeding sounds will be processed as -Qi words. The decision is made with due consideration.

FIG. 1 is a time chart for explaining the operating principle of the present invention, and FIG. 1(a) is an example of the average signal level of the input signal, where A is the input sound t1 signal, and T1 and T2 are the threshold values. This is the point at which vl changes. Figure 1(b) is similar to Figure 1(a).
) is the voice section signal detected using the threshold value. Also, t5
FIG. 1(c) is a voice section end pulse that is generated when the section signal is low after a predetermined time t (for example, 350 ms) has elapsed since the fall of the voice section signal in FIG. 1(b). As will be described later, a sample hold circuit operates and the threshold value is switched.

FIG. 2 is an electrical block diagram for explaining one embodiment of the voice section detection circuit according to the present invention. In the figure, ■ is an input terminal, 2 is a detection circuit, 3 is a smoothing circuit, and 4 is a sample hold. 5 is a level comparison circuit, 6 is a voice section detection circuit, 7 is an output terminal, the average level of the input signal from the input terminal 1 is detected through the detection circuit 2 and the smoothing circuit 3, and the average level signal is sampled and held. The signal is input to the circuit 4 and the level comparison circuit 5. The output signal from the level comparison circuit 5 is input to the voice section detection circuit 6, and the voice section end pulse generated here is input to the sample hold circuit 4.
The signal level at that time is held. This held value and the average signal level are compared in the level comparison circuit 5.
Then, the voice section signal is output to the output terminal 7. Hereinafter, the voice end pulse level is used as is for the next rll Conceivable.

FIG. 3 is a diagram showing an example of a voice section detection circuit constructed based on the above-mentioned idea. In the figure, 8 is a reference voltage generation circuit for generating a constant bell signal Vs, and 9 is an adder circuit. The parts having the same functions as those in FIG. 2 are designated by the same reference numerals as in FIG. 2. In this embodiment, the output signal of the sample and hold circuit 4 is supplied with a constant bell signal Vs from the reference voltage generation circuit 8.
is added, and this added (f+ is the new Ill 4
fi, which makes it possible to deal with noise that is 17 times louder.

Operation--1 As is clear from the above description, according to the present invention, it becomes possible to detect a section signal by determining an optimum value, and it becomes possible to perform speech recognition more accurately.

[Brief explanation of the drawing]

FIG. 1 is a time chart for explaining the operating principle of the present invention, and FIGS. 2 and 3 are electrical block diagrams for explaining the present invention in detail. Terminal, 2... Detection circuit, 3... Smoothing circuit, 4... Sample hold circuit, 5... Level comparison circuit, 6... Voice section detection circuit, 7... Output terminal, 8... ...Reference voltage source, 9...Addition circuit. Procedural Amendment (Ki) June 6, 1980 Patent Application No. 73538 2, Name of the invention Vocal section detection circuit 3, Relationship with the case of the person making the amendment Patent applicant Ota Ri Nakamagome Address Ota, Tokyo 1-3-6 Nakamagome, Ward Name
(674) Ricoh Co., Ltd. Representative: Hama 1) Hiro 4, Agent Address: 807 No. 6, Sha Train Yokohama, 1-2-7 Furo-cho, Naka-ku, Yokohama 231, Detailed description of the invention in the specification subject to amendment Column 7, Correction details (1), from page 4, line 20 of the specification to page 5, line 1, change ``the pulse level to the next one as it is'' to ``to change the signal level at the time of pulse generation to the next one without changing it''. ”. (2) Correct the "slightly louder noise" described in the second line of page 5 to "larger noise." (3) Correct "thereby a little louder noise" written in the 14th line to "thereby a louder noise".

Claims (2)

[Claims] (1) In a voice section detection circuit that extracts the voice signal power and sets a threshold value within the silent section to cut out the voice section, a pulse is emitted once a predetermined time has elapsed from the end of voice I. and a sample-and-hold circuit that samples and holds the noise level at the time of the pulse generation, and uses the hold value as a sound section extraction threshold? A voice section detection circuit characterized by detecting a \rjH section. (2), having one stage for adding a certain fixed value to the threshold;
The speech section detection circuit according to claim 1, wherein the speech section is detected by using the value as a value for cutting out the speech section.