JPS61183699A - Voice recognition equipment - 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 Field of the Invention The present invention relates to a speech recognition device that recognizes input speech using standard patterns of speech registered in advance.

2. Description of the Related Art In recent years, voice has attracted attention as an input means for human-machine systems. It is also being considered that sentences can be created using voice recognition. In speech recognition, the standard pattern configuration is a major factor in determining the performance of the device.

It is known from ``Real-time Recognition Device for Monosyllabic Speech Transmissions'' (Tatsu Ifukube, Nei Nitori 1 et al.) Journal of the Acoustical Society of Japan (1983) 39-2.

FIG. 3 shows the configuration of this conventional speech recognition device, which will be explained below.

3 is an A/D converter that digitizes the electrical signal from the A/D converter 2; 3 is a feature extractor that extracts audio features from the audio signal digitized by the A/D converter 2 by means such as frequency analysis; 4 is a feature extractor that This is a vowel detection unit that detects the vowel of the voice extracted by the extraction unit 3. 6 is a vowel input pattern storage unit that temporarily stores the vowel part pattern detected by the vowel detection unit 4 as a vowel input pattern; 6 is a vowel input pattern storage unit that temporarily stores the vowel part pattern detected by the vowel detection unit 4; A vowel standard pattern storage section 7 stores a vowel standard pattern that is determined to be most similar to the vowel input pattern from among the plurality of vowel standard patterns stored in the vowel standard pattern storage section 6. A recognition unit that recognizes the input voice as a result of vowel recognition based on the recognition result and the characteristics of the voice other than the vowel, and an output unit that outputs a signal corresponding to the voice recognized by the recognition unit 7. It is.

The operation of the speech recognition device configured as described above will be specifically explained below.

When audio is input to the input unit 1, the A/D conversion unit 2 digitizes it, the feature extraction unit 3 extracts the features of the digitized audio signal, and the vowel detection unit 4 extracts the features of the digitized audio signal. The recognition unit 6 determines that the pattern is the most similar from among the vowel standard pattern storage unit which has one vowel standard pattern corresponding to the vowel category as shown in FIG. 4 which has been registered in advance. The signal corresponding to the vowel standard pattern is used as the vowel recognition result, and based on the recognition result, the input speech is recognized with features of speech other than vowels, and a signal corresponding to the recognized speech is output from the output unit 1°1. .

Problems to be Solved by the Invention The vowel standard pattern storage section of the speech recognition device configured as described above has a drawback in that it cannot cope with variations in vowel parts. For example, if you say Ital, I-
1 is slightly different from the 1-1 pattern uttered as a single vowel, even if it is the same Ial pattern due to the influence of the consonant part. In the conventional vowel standard pattern configuration, one vowel is uttered for each vowel category. However, since only the standard vowel pattern exists,
In cases like the previous example, this may cause erroneous recognition of other vowels. As described above, the vowel standard pattern storage section that has only one vowel standard pattern corresponding to each vowel category has a problem in that it cannot cope with variations in the vowel part of each syllable.

In view of this, an object of the present invention is to provide a speech recognition device that reduces vowel errors due to vowel variations and has a high recognition rate.

Means for Solving the Problems In order to solve the above problems, the present invention provides a speech recognition device including a plurality of vowel standard pattern storage sections each having a plurality of vowel standard patterns corresponding to each vowel category. be.

Effect of the Invention The present invention has a plurality of standard patterns corresponding to each vowel category due to the configuration of the vowel standard pattern storage section, which expands the range of absorbing vowel fluctuations due to the influence of consonants. Even if part 1 is not recognized as laI of a monophthong, if it has a high degree of similarity to lal of a voice with a vowel part of l-1 influenced by a consonant part, a correct recognition result as l-1 can be obtained. It is possible to improve the recognition rate.

Embodiment FIG. 1 shows a block diagram of a speech recognition apparatus according to an embodiment of the present invention, and this will be explained.

In the figure, 1 is an input section consisting of a microphone etc. that converts audio into an electrical signal, 2 is an A/D converter that digitizes the electrical signal from the input section 1, and 3 is an A/D converter 2 that converts the electrical signal into a digital signal. This is a vowel detection unit that detects vowels in converted speech. 5 is a vowel manual pattern storage section that temporarily stores the vowel part pattern separated by the vowel detection section 4 as a vowel input pattern; 9-1.9-2.・・・・・・
... 9-n is a vowel standard pattern storage unit that stores in advance the characteristics of a plurality of vowels to be recognized as standard patterns, and 9-1.9-2. ...9-n is a standard pattern storage unit for each category of vowels, and n represents the number of categories of vowels.

Figure 2 shows the vowel standard pattern storage section 9-1.9-2°...
...9-n is a configuration diagram. a1. a2,
. . . is the standard pattern 11 of l-1, respectively. *
12, i3. . . . are the standard patterns u1 of fil, and u2 . . . are the standard patterns e1 of lul, respectively. θ2, . . . are standard patterns of +e+, and 01.02, 03, . . . are standard patterns of lol, respectively. In this example, n in the vowel standard pattern storage section 9-n in FIG. 1 is 6. In this way, the vowel standard pattern storage section has a plurality of standard patterns corresponding to each vowel category. 7 is determined to be most similar to the vowel input pattern from among the plurality of vowel standard patterns stored in the vowel standard pattern storage unit 9-1.9-2...9-n. a recognition unit that recognizes a signal for a standard vowel pattern as a vowel recognition result, and recognizes input speech with features of speech other than vowels based on the recognition result;
Reference numeral 8 denotes an output section that outputs a signal corresponding to the voice recognized by the recognition section.

The operation of the speech recognition device configured as described above will be specifically explained below.

First, when a desired voice is inputted into the input section 1, the pattern detected by the vowel detection section 4 corresponds to the vowel category registered in advance by the recognition section 7.
The signal corresponding to the vowel standard pattern that is determined to be most similar to the vowel input pattern from among the plural vowel standard patterns is taken as the vowel recognition result, and based on the recognition result, the input voice is combined with the features of the voice other than the vowel. and outputs a signal corresponding to the recognized voice from the output unit 8.

As described above, according to this embodiment, by providing a plurality of standard vowel patterns corresponding to each vowel category, the vowel manual pattern changes due to the influence of consonants,
Even if the similarity to monophthong patterns is low, correct recognition results can be obtained by finding patterns with high similarity among other standard vowel patterns influenced by consonants. By improving the vowel recognition rate in this way, the recognition rate of this speech recognition device can be improved.

In the embodiment, the vowel standard pattern storage section 9-1.
9-2...9-n represents the number of vowel categories and is variable.

As described in detail, according to the present invention, by having a plurality of vowel standard patterns corresponding to each vowel category,
Even if a change occurs in the vowel part of the voice, vowel recognition can be performed accurately, resulting in a high recognition rate, which has a great effect.

[Brief explanation of drawings]

Fig. 1 is a block diagram of a monosyllable recognition device according to an embodiment of the present invention, Fig. 2 is a block diagram of a vowel standard pattern storage unit of the monosyllable recognition device of the same embodiment, and Fig. 3 is a block diagram of a conventional monosyllable recognition device. FIG. 4, a block diagram of the device, is a layout diagram of a vowel standard pattern storage section of a conventional monosyllable recognition device. 1... Input section, 2... A/D conversion section,
3... Feature extractor, 4... Vowel detector, 5... Vowel input pattern storage unit, 6...
・Conventional vowel standard pattern storage unit, 79. ...Recognition section, 8...Output section, 9-1 to 9-n...
- Vowel standard pattern storage unit of the example. Name of agent: Patent attorney Toshio Nakao and 1 other person No. 3
figure

Claims (1)

[Claims] a feature extractor that extracts the features of the input speech; a vowel detector that detects the vowels of the feature-extracted speech; and a vowel part pattern detected by the vowel detector that is stored as a vowel input pattern. a vowel input pattern storage unit; a plurality of vowel standard pattern storage units that store vowel part patterns detected by the vowel detection unit as a plurality of vowel standard patterns corresponding to each vowel category; The signal for the vowel standard pattern that is determined to be most similar to the vowel input pattern from among the standard patterns is taken as the vowel recognition result, and based on the recognition result, the input speech is recognized using features of speech other than vowels. A speech recognition device comprising: a recognition section.