JPS60149098A - Voice input unit - 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 l(+ Field of Industrial Use) The present invention relates to a voice input device for voice inputting kana characters by recognizing voice.

10) Prior Art This persimmon speech input device is generally configured to adopt a monosyllabic input method for all kinds of entrance speeches, and to be able to recognize all monosyllabic speech.

In the case of such a monosyllabic input device, for example, when inputting the word “T5KYδ”, the operator inputs “T6”, “KYOJ” and l1l(j ?
If the kana string ``Tokyo'' is input as LJ l°2, it does not mean that the kana string ``Tokyo'' is exactly the same.

In this way, in conventional devices, in order to represent kana strings of long syllables such as ``to'' and ``Fkyo,'' for example, after vocalizing ``TO,'' you manually add ``-'', which is a sound display symbol, to 1. Since input operations had to be performed, complete voice input without manual input could not be achieved.

OBJECT OF THE INVENTION The present invention provides a voice input device that automatically recognizes the input syllable as a long syllable even when inputting a long H syllable, making it possible to input the long syllable without manual operation. Deacchi.

B) 1 Structure of the Invention The speech input device of the present invention detects the duration of a vowel present in each syllable of input speech, and when this duration exceeds a predetermined time, determines that the syllable is a long syllable. The apparatus is provided with a long syllable detecting means for detecting a long syllable, and automatically inputs a long syllable using the long syllable detecting means.

Thu) Example The figure shows an embodiment of the voice input device of the present invention.

In the figure, (11 is a microphone that converts voice into an electrical signal, and (2) is a microphone that extracts characteristic parameters such as frequency spectrum values from the syllable-based voice signal obtained by the microphone (1). This is a stone parameter extraction circuit. (3) is a reference parameter memory that stores standard feature parameters for all syllables as reference parameters in advance. (4) is the reference parameter memory + 31 & 6 Soariru. It is a recognition circuit that compares and recognizes all the unknown parameters of the unknown syllable obtained from the parameter extraction circuit (21) based on each associated parameter, and outputs the syllable information indicated by the reference parameter that has the minimum error with the unknown parameter. (5), and (6
) is the katakana code for displaying line information and the syllable,
and a katakana table as a kana code generation circuit that converts it into a flat name code, and a flat name table. table (5),
(6) is selectively input. The switching operation of the kana changeover switch means (81) is performed by operating the kana designation key (K).

On the other hand, (7) is a vowel detection circuit that detects the vowel part at the rear end of the unknown parameter in syllable units obtained from the parameter extraction circuit (2), and when this unknown parameter is a frequency spectrum value, A vowel part whose spectral value changes over time can be distinguished from a consonant part whose spectral value changes over time and can be extracted. (8) is a counter that counts the duration of the vowel detected by the vowel detection circuit (7), (9) is a register that stores the threshold for the duration of the vowel in a long syllable, and This is a comparator that compares the count value of the counter (8) with the threshold value of the register (9), and outputs a long soup syllable detection signal when the count value of the counter IRI exceeds the threshold value of the register (9).

The old) is a katakana long syllable code addition circuit that receives the long syllable detection means from the comparison circuit rUQ and adds a long code indicating a long syllable, that is, "-", following the katakana code obtained from the kikatakana table (5). It is.

11z is a hiragana long vowel code addition circuit which receives the long vowel detection signal from the comparison circuit section and adds a flat name code indicating a long syllable following the flat name code obtained from the long vowel name table (6). For example, after "to", the flat name code "U" is output. 1131 is the above-mentioned double major chord addition circuit +111. CR that displays and outputs characters based on katakana from fj2 or flat name code
It is T.

Therefore, the voice input device of the present invention has a kana designation key (K1
Kana changeover switch means 181 by operating
When connected to the katakana table (5) side, the mode becomes katakana mode, and conversely, when the means TS+ is connected to the plain name table (6) side, the mode becomes the plain name mode fj.

In such katakana mode, in order to input, for example, the voice "Tokyo" into the microphone (1), "TO" is input.

This is followed by inputting ``°KYO'', and these speech signals for each syllable are extracted as parameters (2) and recognized (4), and this recognition result is converted into a katakana table (5). It is often converted into the codes of "to" and "kiyo". At this time, the vowel detection circuit (7) detects the instrument part of °0 from the voice signal -Too, and the counter (8)
counts the duration of this vowel, comparator ttO1 detects that the duration of this vowel is greater than the threshold of register (9), and operates the katakana long code addition circuit (Ill). The circuit (Ill outputs the code "-" following the "to" code from the katakana table (51), and then outputs the "-" code following the "kiyo" code.
Since there is a full output, 0RTO31 displays the katakana characters "To" and "Kiyo".

On the other hand, in flat name mode, microphone (1)
When "Tδ" - KYδ° is input to , the hiragana table (6) is used to obtain the codes for "to" and "kiyo", but in the same way as in the katakana mode described above,
Since the long syllable detection signal activates the hiragana long syllable adding circuit (6), 0RT031jC is displayed as "tori" and "kiyo".

g) Effects of the invention As is clear from the above description, the voice input device of the present invention has the following effects:
A long syllable detection means is provided for detecting the duration of a vowel present in each dorsal syllable of the input speech, and detecting that the vowel is a long syllable when the duration of the vowel exceeds a predetermined time; Since the long sound display code is added to the original kana display code when the long sound syllable input is detected, the long sound syllable can be automatically recognized and complete speech input without manual operation is possible.

[Brief explanation of the drawing]

The figure is a block diagram showing the configuration of an embodiment of the voice input device of the present invention, (2) is a parameter extraction circuit, (4) is a recognition circuit, +51f6+ is a kana table, (7) is a vowel detection circuit, ( 8) is a callan, (9) is a register, a
nt indicates a comparison circuit, and OH3 indicates a long code addition circuit. Applicant Sanyo Electric Co., Ltd. Representative Patent Attorney Shizuo Sano

Claims (1)

[Claims] 1) Comparison of a parameter extraction circuit that extracts voice characteristic parameters from a voice signal in units of back parts, and monosyllabic parameters obtained from the extraction circuit based on a plurality of reference monosyllabic parameters stored in advance in memory. A recognition circuit that recognizes a recognition circuit, a kana code generation circuit that generates a kana code indicating the recognition result based on the recognition result obtained from the recognition circuit, and a kana code generation circuit that detects the duration of a vowel present in each syllable of the audio signal. , a master syllable detection means for detecting that the syllable is a long syllable when the duration exceeds a predetermined time period; and a master syllable detection means for detecting that the syllable is a long syllable when the duration exceeds a predetermined time period; a long code adding circuit for adding a long code to the kana code generated, and display means for receiving each display code string obtained by the kana code generating circuit and the long code adding circuit and displaying the kana. input device.