JPS60225271A - Kana-kanji converting device of voice input - Google Patents

Abstract

PURPOSE:To shorten the voice recognition and conversion processing time by using breath sounds and information of intervals, which are inputted at a voice input time, to cut efficiently a straight inputted Kana (Japanese syllabary) character string to clause units. CONSTITUTION:An input voice waveform is inputted to a voice input part 1, and feature parameters are extracted by a voice analyzing part 7 in the voice input part 1. The voice waveform is recognized in monosyllabic units by a phoneme recognizing part 9, and an interval detecting part 10 detects intervals, and a breath sound detecting part 8 detect breath sounds B, and these detection results are transmitted to a central control part 4 through a voice part controller 11. The central control part 4 discriminates the code (based on detection of breath sounds and detection of intervals), by which the input Kana character string is divided to clause units, from the voice part controller 11. When this code is transmitted, a Kana-Kanji (Chinese character) converting part 5 is started to perform the straight input Kana-Kanji conversion processing of the Kana character string which is inputted with a voice and is recognized.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to an improvement of a voice input kana-kanji conversion device.1. More specifically, the present invention relates to a speech input kana-kanji conversion device that can efficiently perform kana-kanji conversion processing using information on breath sounds and sound spaces (ma) in input speech.

<Technical background of the invention and its problems> In general, it is not necessary to input additional information such as bunsetsu delimiter marks, and the so-called so-called "sentence" that automatically converts and outputs a sentence containing kana and kanji by inputting the pronunciation as per the pronunciation. In a solid input kana-kanji conversion device, it is necessary to determine the clause boundaries of an input sentence.

In conventional kana-kanji conversion devices that use solid input, segmentation of phrases is performed using grammatical analysis, etc.
It took time to extract each clause, making it difficult to process in real time.

<Object of the Invention> The present invention eliminates the above-mentioned conventional problems and makes it possible to efficiently cut out phrases by using the press sound of each phrase and the information between the punctuation mark units when inputting voice. An object of the present invention is to provide a voice input kana-kanji conversion device, and in order to achieve this purpose, the voice input kana-kanji conversion device of the present invention includes a detection means for detecting press sounds and gaps in input voice, and this detection means. and a kana-to-kanji conversion means for converting a character string of input speech into kana-to-kanji characters using the detected output as clause delimiter information.

<Embodiment of the Invention> Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a voice input kana-kanji conversion device to which the present invention is applied.

In Figure 1, %1 is the voice input section, 2 is the key manual section,
3 is a display unit, 4 is a central control unit, 5 is a kana-kanji conversion unit, 6
is a kana-kanji conversion dictionary storage unit, and the central control unit 4 is constituted by, for example, a microcomputer, and controls the other components 1 to 3 and 5 to 6, and also controls the other components 1 to 3 and 5 to 6. The system is configured to use the computer to perform various processes that will be described later. The dictionary storage unit 6 also includes dictionaries necessary for segmentation processing, homophone selection processing, kana-kanji conversion processing, and the like.

The kana-kanji conversion section 5 uses the contents of the kana-kanji conversion dictionary storage section to input letters in units of compound phrases separated between press sounds and sounds, which will be described later. A kana-kanji conversion process is performed, and the conversion result is displayed on the display unit 3.

FIG. 2 is a block diagram showing a detailed configuration example of the audio input section l, and FIG. 3 is a diagram for explaining its operation.

In FIG. 2, reference numeral 7 indicates a voice analysis unit that extracts sounds detected by a microphone or the like, and 8 indicates whether or not the input voice is a press sound based on the characteristic parameters of the input voice analyzed by the voice analysis unit 7. a press sound detection unit for detecting 9
is a phoneme that is recognized for each phoneme (single syllable) by performing a matching process between the feature parameters of the input speech that has been speech-analyzed by the speech analysis unit 7 and a pre-stored standard feature pattern of phonemes (single syllables). (monosyllabic) recognition section 10 is reset by the input speech from the speech analysis section 7, and an input speech interval detection section 11 is provided with a counter started on the basis of a signal from the speech section control device 11. The press sound and the detection signal between the sounds are sent to the central control unit 4 along with the phoneme (monosyllable) information, and an instruction signal is sent to the central control unit 4 to activate the counter in the pause detection unit 10 in response to the input of the press sound and the phoneme signal. This is an audio section control device for deriving.

In the above configuration, continuous input speech is input to the speech analysis section 7, feature parameters of the input speech are sequentially extracted with respect to the input time t, and the input speech analyzed into the feature parameters is sent to the press sound detection section 8 and the phoneme recognition section 8. The signal is sent to the interval detection section 9 and the interval detection section IO.

The press sound detection section 8 determines whether or not the received input sound is a press sound based on characteristic parameters such as the analyzed power p(1), and if it is a press sound, the sound section control is performed. Device 111C notifies that a press sound has been input, and upon receiving this signal (press sound detection signal), the audio section control device 11 centrally controls a code signal for dividing into compound phrase units in the above-mentioned kana-kanji conversion section 5. It is derived in Part 4.

The phoneme recognition unit 9 performs matching processing between the voice characteristic parameters of the input voice and the standard phoneme characteristic pattern, and if there is a corresponding phoneme, it sends the phoneme code to the voice unit control device 11, and this phoneme code is The audio unit control device 11 that received the phoneme code directly sends the phoneme code to the central control unit 4.

The pause detection section 10 resets a counter in the pause detection section IO in response to the voice input via the speech analysis section 7. Further, the audio section control device 11 includes a press sound detection section 8.
When receiving the press sound detection signal from the phoneme recognition section 9 or the phoneme code from the phoneme recognition section 9, the amount detection section 10C sends an instruction signal to start the counting operation of the counter in the pause detection section 10C.

During this time, if the count value of the counter in the detection unit 10 exceeds a certain value, it is determined that there is a pause, and the pause detection unit 10 transmits the presence of a pause to the audio unit control device 11 as a pause detection signal. do. During this time, the speech section control device 11 receiving the detection signal outputs a code signal to the central control section 4 for dividing into compound phrase units in the above-mentioned kana-kanji conversion section 5.

Now, when the input speech waveform ■ shown in FIG. The recognized character string R "Konkai no Shutsugan..." as shown in FIG. In addition, the amount detection unit IO detects the pause (
m) S is detected, and the detection result is sent to the central control section 4 via the audio section control device 11. Further, the press sound detection section 8 detects the press sound B after the syllable "n", and the detection result is sent to the central control section 4 via the voice section control device II. Therefore, in the central control unit 4, the detection signal for the character string and press sound detection 1 is
...'' will be input in this order.

The central control unit 4 determines whether or not a code (signal) for dividing into phrases is sent from the audio section control device 11 (based on detection of press sound 9), and sends a code for dividing into phrases. When the text is input, the Kana-Kanji conversion unit 5 is activated to perform the solid input Kana-Kanji conversion process on the kana character strings that have been voice input and recognized. In order to cut out a kana character string from a compound clause unit to a single clause unit, clause division processing is performed while referring to the dictionary, table, etc. necessary for the division process of cutting out the single clause unit stored in the kana-kanji conversion dictionary storage unit 6. After that, the kana-kanji conversion processing unit 5 performs kana-kanji conversion processing for each clause, and checks the context with the previous clause. The phrase break is removed and the previous conversion is canceled.If the phrase can be properly extracted and there is no contradiction in the context of the phrase, the conversion result is output to the display unit 3 or the like.

Note that, since the bunsetsu segmentation process and the kana-kanji conversion process themselves are conventionally known processing techniques, detailed explanation thereof will be omitted.

As described above, kana-kanji conversion processing is performed by rapidly cutting out phrases from the kana string of the input speech recognition results using press sounds or ma (ma) information entered in phrases or punctuation marks. . In addition, by activating the kana-kanji conversion process by detecting the press sound or between the sounds (ma), the conversion result can be output at any time without waiting for the end of inputting each sentence. The conversion process is performed without the need for

<Effects of the Invention> As described above, according to the present invention, there is provided a detection means for detecting press sounds and sounds in an input voice, and a kana name of a character string of the input voice using the detection output of the detection means as phrase delimitation information. Since it is equipped with a kana-kanji conversion means that performs kanji conversion, it is possible to efficiently convert kana character strings input in solid text into phrases using the information of the press sound 2 ma (ma) input during voice input. As a result, it becomes possible to shorten the speech recognition and conversion processing time.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a voice input kana-kanji conversion device to which the present invention is applied, FIG. 2 is a block diagram showing an example of the configuration of the voice input section, and FIG. 3 is a diagram for explaining its operation. It is. 1... Voice input section, 3... Display section, 4... Central processing section, 5... Kana-kanji conversion section, 6... Kana-kanji conversion dictionary storage section, 7... Speech analysis section , 8... Press sound detection section, 9... Phoneme (single syllable) recognition section, lO... Amount detection section.

Claims (1)

[Scope of Claims] 1. Detection means for detecting press sounds and gaps in input speech, and kana-kanji conversion for converting character strings of input speech into kana-kanji using the detection output of the detection means as clause delimiter information. A voice input kana-kanji conversion device characterized by comprising: means; 2. The kana-kanji conversion means is configured to activate the kana-kanji conversion of the character string input so far in response to the detection output of the detection means. voice input kana-kanji conversion device.