JPS61277998A - Voice recognition 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] [Industrial application field] The present invention relates to a speech recognition circuit that performs speech recognition.

[Conventional technology]

Various methods have been proposed for speech recognition circuits. An example of this is the speech recognition circuit shown in FIG.

In FIG. 3, reference numeral 1 denotes a speech analysis section, which analyzes the input speech input signal S (for example, spectrum analysis) for each fixed analysis section, and: - determines characteristic parameters and outputs an analysis result output A; A recognition unit 2 matches the input analysis result output A with a standard pattern registered in advance, and outputs the degree of difference (distance) as a recognition result output M.

FIG. 4 shows an example of a typical audio signal waveform. T
is one analysis section, and *T1 indicates a section of the audio signal where short sections such as plosive sounds occur.

[Problem that the invention seeks to solve]

However, in conventional circuits, T is generally set to be longer than T in order to perform stable analysis, and the original features of vocal sections where a steady state continues for a long time, such as vowels, are extracted. For parts such as plosive sounds that occur only in short intervals, there is a problem in that the features are not sufficiently extracted at the ends of the analysis interval.

SUMMARY OF THE INVENTION An object of the present invention is to provide a speech recognition circuit that can always extract the characteristics of a short speech interval such as a plosive using amplitude information.

[Failure to solve the problem]

According to the present invention, there is provided an analysis section that analyzes an input signal representing speech for each predetermined analysis section and outputs the results.

The speech recognition circuit includes a recognition section that recognizes input speech based on the analysis result output from the analysis section, an amplitude detection circuit that detects the amplitude of the input signal, and an amplitude detection circuit that detects the amplitude of the input signal based on the output of the amplitude detection circuit. There is obtained a speech recognition circuit characterized in that it includes a synchronization section that changes the temporal position of the analysis section of the analysis section.

〔Example〕

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

FIG. 1 is a schematic diagram of one embodiment of the present invention.

The difference from the conventional block diagram in FIG. 3 is that an amplitude detection circuit 3 for detecting the start of a plosive, etc.;
The point is that a synchronization section 4 is provided which changes the frame position to be analyzed (that is, the temporal position of one analysis section) based on the output of.

The analysis interval changing operation in this embodiment will be explained with reference to FIG. FIG. 2(a) is the Fi input audio signal waveform, and the tn point is the starting point of a plosive. The amplitude detection circuit 3 receives the audio signal shown in FIG. Based on the information X that detected the starting point B of the plosive, the synchronization unit 4 performs the same process as previously performed in the analysis section T of FIG. 2(b).

The temporal position of the analysis section of the speech analysis section 1 is changed so that the starting point B is located at the center of the analysis section as shown in FIG. 2(c).

This can be realized using numerical elements. In addition, the synchronization section 4
This can be achieved by having twice the buffer memory and determining the starting end of the analysis section on the buffer memory so that the starting point B is set at the center of the analysis section.

〔Effect of the invention〕

As explained above, in the present invention, it is possible to correctly extract the characteristics of plosives, etc. by including an amplitude detection circuit that detects the starting point of plosives, etc., and a synchronization section that changes the two analysis intervals. The effect of improving speech recognition performance is significant.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a speech recognition circuit according to an embodiment of the present invention, FIG. 2 is a diagram for explaining the analysis interval changing operation in an embodiment of the present invention, and FIG. 3 is a conventional speech recognition circuit. FIG. 4 is a block diagram showing typical audio signal waveforms. (C)＼゛Figure 4-11 A-11

Claims (1)

[Claims] 1. Speech recognition having an analysis section that analyzes an input signal representing speech for each fixed analysis section and outputs the results, and a recognition section that recognizes the input speech based on the analysis results output from the analysis section. A speech recognition circuit comprising: an amplitude detection circuit that detects the amplitude of the input signal; and a synchronization section that changes the temporal position of the analysis section of the analysis section based on the output of the amplitude detection circuit. .