JPS63300297A - 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] [Industrial application field] The present invention relates to a speech recognition device.

[Prior Art] For example, in a conventional speech recognition word processor, all speech human input information is used as word recognition candidates.

[Problems to be Solved by the Invention] Therefore, in such a speech recognition word processor, for example, in response to a speech input of /fail/ (file), if it can respond to an external sound, it will "enter" in the recognition result. Confusion such as "file" could occur. In addition, for all human voice information, words that require kanji conversion,
Foreign Words 1 Since the usual Japanese hiragana, new words, onomatopoeia, etc. are all used as candidate words, processing is extremely wasteful.

SUMMARY OF THE INVENTION An object of the present invention is to provide a speech recognition device that eliminates the above-mentioned drawbacks, speeds up recognition processing by selecting recognition candidates, and enables manual recognition of new words and onomatopoeia.

[Means for Solving the Problems] The present invention provides a voice recognition means for recognizing human voice information by comparing it with at least one standard pattern information, and combining the output information of the voice recognition means with dictionary information as necessary. a word matching means for obtaining word recognition information by matching; an input word specifying means for specifying an input word; and an input word specifying means for specifying an input word; and selection means for selecting universal dictionary information in the matching means.

[Function] According to the present invention, based on the designation of the input word designation means, standard pattern information to be applied in the speech recognition means is selected, and, if necessary, dictionary information to be applied to the word matching means is selected.

[Example] In the present invention, it does not matter whether the human voice is a single syllable or a continuous word. Furthermore, the standard pattern
It does not matter whether it is a monosyllable, vCv, phoneme, etc. In the following embodiment, an example will be shown in which human speech is recognized using phonemes.

FIG. 1 is a block diagram of an embodiment of the present invention applied to a speech recognition word processor. 1 is an input speech signal, which is manually input to the phoneme recognition circuit 2.

3 is an input word designation signal, which is output from the keyboard when an input word designation key on the keyboard is pressed. Reference numeral 4 designates a first switch which determines whether or not to add the output signal of the extraneous sound standard pattern circuit 6 to the output signal of the standard pattern circuit 5 in addition to the output signal of the standard pattern circuit 5, based on the input word designation signal 3. select. Reference numeral 7 denotes a word matching circuit, which performs matching between the output signal from the phoneme recognition circuit 2 and each dictionary output described below, recognizes a word with respect to the signal from the phoneme recognition circuit 2, and outputs a recognition result 12. Reference numeral 8 denotes a second switch which, based on the human input signal 3, selects a kanji dictionary (word dictionary that uses kanji in Japanese) 9, a Japanese hiragana dictionary 10, and a foreign word dictionary (usually uses katakana) 11. It is selected whether or not to give the output of any one of them to the word matching circuit 7.

Next, the switching operations of the first switch 4 and the second switch 8 based on the input word designation signal 3 will be explained.

First, when the input word designation signal 3 is a foreign word designation signal (specifically, for example, designates katakana input on the keyboard), the first switch 4 closes and the foreign sound standard pattern circuit 6 is selected. At this time, in the second switch 8, the switch 2-3 is closed to select (close) the foreign word dictionary 11, and the other switches 2-1 and 2-2 are opened.

When the input word designation signal is a Kanji conversion designation signal, the first switch 4 is opened. At this time, the second switch 8 is the switch 2-1.
closes and selects Kanji dictionary 9 (other switches 2-2.
2-3 open). When the input word designation signal 3 is a hiragana designation signal, the first switch 4 is opened and the second switch 8, switch 2-2, is closed to select the Japanese hiragana dictionary 10 (other switches 2-1, 2-2, etc.) 3 is open). When the input word designation signal 3 is a non-conversion designation signal and a hiragana or katakana designation signal, the first switch 4 is closed and the external sound standard pattern circuit 6 is selected, and the second switch 8 is all opened and the dictionary is not used. . At this time, regarding the signal from the phoneme recognition circuit 2,
In the word matching circuit, hiragana is converted into hiragana and katakana is converted into katakana, and a recognition result 12 is output.

FIG. 2 is a block diagram of another embodiment of the present invention applicable to a speech recognition word processor. As shown in the figure, the phoneme recognition circuit 21 compares the input voice information 1 with the normal standard pattern and the extraneous sound pattern 56, and the phoneme selection circuit 41 outputs the input word designation signal 3 with respect to the output of the circuit 21.
The output of the same circuit 41 is sent to a word matching circuit 71 to select a phoneme based on the kanji, hiragana, and foreign word dictionary 91.
Based on the input word designation signal 3, the word selection circuit 81 selects a word based on the input word designation signal 3 based on the output of the matching circuit 71. Even with such a configuration, extremely efficient speech recognition is possible.

[Effects of the Invention] As explained above, according to the present invention, it is possible to eliminate word confusion, and furthermore, it is possible to speed up processing by specifying words in a dictionary. In addition, it also enables human input regarding unregistered words.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of one embodiment of the present invention, and FIG. 2 is a block diagram of the same embodiment. 4...First switch, 8...Second switch. m1 figure 2

Claims (1)

[Scope of Claims] 1) Speech recognition means that performs speech recognition by comparing input speech information with at least one standard pattern information; and matching output information of the speech recognition means with dictionary information as necessary to identify words. a word matching means for obtaining recognition information; an input word specifying means for specifying an input word; and, based on the specification of the specifying means, selecting applicable standard pattern information in the speech recognition means and, if necessary, controlling the word matching means. 1. A speech recognition device comprising: selection means for selecting applicable dictionary information. 2) In the speech recognition device according to claim 1, the selection means adds normal standard pattern information and foreign sound standard pattern information to the applied standard pattern information when the input word designation means designates a foreign word. and selecting foreign word dictionary information as applied dictionary information. 3) In the speech recognition device according to claim 1, the selection means selects normal standard pattern information as the applied standard pattern information when the input word specifying means specifies kanji conversion, and selects the normal standard pattern information as the applied standard pattern information, and selects the normal standard pattern information as the applied standard pattern information A speech recognition device that selects kanji dictionary information. 4) In the speech recognition device according to claim 1, when the input word specifying means specifies hiragana, the selection means selects the normal standard pattern information as the applied standard pattern information, and selects the normal standard pattern information as the applied dictionary information. A speech recognition device that selects Japanese hiragana dictionary information. 5) In the speech recognition device according to claim 1, when the input word specifying means specifies hiragana and no conversion or katakana and no conversion, the selection means adds normal standard pattern information to the applied standard pattern information. and a speech recognition device that selects an external sound standard pattern and does not select dictionary information.