JPS6095598A - Voice recognition circuit - Google Patents

Abstract

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

Description

TECHNICAL FIELD The present invention relates to a speech recognition device, and more particularly to a speech recognition circuit used in a speech recognition device that inputs telephone speech.

BACKGROUND OF THE INVENTION Various methods have been proposed for voice recognition devices, and they have also been applied to cases where the input signal is telephone voice. As an example of this, a block diagram of a conventional speech recognition circuit is shown in FIG.

Referring to FIG. 1, reference numeral 1 denotes a voice analysis section, which analyzes an input voice input signal S (for example).

Spectrum analysis) is performed, the characteristic parameters are determined, and the analysis result output A is output. Reference numeral 2 denotes a matching unit that matches the input analysis result output A with a standard pattern registered in advance, and outputs the degree of difference (distance) as a matching result output M. Reference numeral 3 denotes a determination unit, which uses the input matching result output M to output the category of the pattern with the smallest distance as a determination result output. Conventionally, this judgment result output was referred to as the recognition result output R.

Here, when the voice input signal S is a voice input signal of a telephone line, it is convenient to also recognize signals other than voice, that is, signals specific to the telephone line (e.g., busy tone, ringback tone, etc.). is good.

FIG. 2 shows a typical waveform example of a telephone line signal. Taking a busy tone as an example of a signal unique to a telephone line, the signal (400Hz sine wave signal) transmission section T1 is 500m5ec, and the no-signal section T2 is the same <500m.
5ec, and these alternating repetitions continue. A busy tone is a signal that is sent when the other party on the other end of the phone puts the handset of the telephone on-hook, and when the voice recognition device detects this busy tone, it disconnects one line. Therefore, detection of this busy tone is important and must be done carefully.

However, in conventional circuits, the noise peculiar to telephone lines (
For example, crosstalk, line noise, etc.), this noise may be mistakenly recognized as an important signal such as a busy tone, resulting in 9 lines being mistakenly disconnected.

OBJECTS OF THE INVENTION An object of the present invention is to provide a speech recognition circuit for telephone speech that reduces false recognition and increases reliability by utilizing the periodic repetition of signals unique to telephone lines. be.

Structure of the Invention According to the present invention, there is provided a speech recognition circuit for telephone speech, which includes an analysis section that analyzes an input signal, a pattern output from the analysis section, and a pre-registered speech and telephone signal recognition circuit. a matching unit that performs matching with a standard pattern; a determining unit that selects an optimal value from among the values obtained from the matching unit and outputs a determination result as a recognition candidate; A speech recognition circuit is obtained that includes a counter section that stores the recognition result and outputs the recognition result when the number of times the recognition result reaches a preset number of times as a candidate.

3- Examples of the invention The present invention will be described in detail below with reference to the drawings.

FIG. 5 is a block diagram showing the configuration of an embodiment of the speech recognition circuit according to the present invention, in which a counter section 4 for outputting the Rth signal is added.

FIG. 4 shows a block diagram of the configuration of the counter section 4 shown in FIG. 6 in more detail.

Next, the operation of the present invention will be explained with reference to FIGS. 6 and 4.

The audio input signal S is analyzed by the audio analysis section 1 into several parameters. The matching section 2 inputs the pattern A output from the speech analysis section 1, and matches the input pattern A.
A matching result M is outputted to the determination unit 6. The determination unit 3 outputs the smallest (closest) pattern category 4-1 from the matching result outputs M to the counter unit 4 as a determination result output (recognition candidate output) D. In the counter section 4, the sent recognition candidate outputs are counted and stored in a counter memory 4-1.

In the counter section 4, the counter memory 4-1 counts the recognition candidate outputs for each category, and this counting result (output of the counter memory 4-1) is stored in a predetermined number stored in advance in the setting memory 4-2. When the value is reached, the comparison unit 4-3 outputs the category as a recognition result output R.

In other words, for categories that have a strong influence on dogs, such as signals that occur repeatedly such as busy tones, or signals that terminate the recognition operation depending on the judgment result, such categories are input as recognition candidate outputs. In this case, instead of outputting the recognition result output R immediately as in the conventional case, the recognition result output R is output when the candidate is input as a candidate twice, for example.

The matching section 22, the determining section 6, and the comparing section 4-3.

Both can be configured with a simple digital circuit or a microprocessor, and can also be configured with a counter memory 4-
1 and setting memory 4-2 can be sufficiently constituted by one ordinary memory or register.

Effects of the InventionAs is clear from the above explanation, the speech recognition circuit according to the present invention reduces false recognition and increases reliability by outputting the recognition result when the number of candidates reaches a preset number. You can increase the degree.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the configuration of a conventional speech recognition circuit, Fig. 2 is a waveform diagram showing an example of a signal peculiar to a typical telephone line, and Fig. 3 is an embodiment of the speech recognition circuit according to the present invention. FIG. 4 is a block diagram showing the counter section 4 shown in FIG. 6 in detail. 1...Speech analysis section, 2...Matching section, 6...
Judgment unit, 4... Counter unit, 4-1... Counter memory. 4-2...Setting memo L4-3...Comparison section.

Claims (1)

[Claims] 1. An analysis section that analyzes the input signal, a matching section that matches the pattern output from the analysis section with a standard pattern that has been registered in advance, and an optimal value from among the values obtained from the matching section. a voice recognition circuit having a determination unit that selects and outputs a determination result, the input signal being a telephone voice, storing the number of times of the determination result;
1. A speech recognition circuit for telephone speech, comprising a counter unit that outputs a recognition result when the number of times the determination result reaches a preset number.