JPH07113837B2 - Character generator - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a character generation device which is used in a word processor or the like and outputs a character sequence.

(Prior Art) Conventionally, there is known a device for creating a sentence by inputting character data from a keyboard or the like and then outputting a predetermined font character string by a printer according to the character data, and a word processor. It is called.

(Problems to be solved by the invention) However, in the conventional word processor, the output character string is uniform and non-verbal information is transmitted without the familiarity as seen in handwritten characters. There was a flaw that was not done.

It is an object of the present invention to provide a character generation device using voice recognition that enables non-verbal information transmission by a document creator and generates a familiar character string.

(Means for Solving Problems) A first invention of the present application is a means for converting an input voice into a character string by performing voice recognition, and a pitch frequency change width or an articulatory coupling degree from the input voice in a non-language. It is composed of means for extracting as data and means for generating characters from the character string and the non-language data.

A second invention of the present application is a means for inputting a character string from a keyboard, a means for extracting non-language data from a voice corresponding to the character string, and a character generated from the character string and the non-language data. And means.

(Principle of the Invention) It is generally known that speech contains non-language information in addition to linguistic information. For example, a change in pitch frequency includes emotional information, and a large change in pitch frequency indicates that the emotion is higher than that in a small change. In addition, the pitch frequency of a word indicating strong intention tends to be high. Further, the voice has a property called articulation coupling, and the pattern of the voice corresponding to the character is influenced by the preceding and succeeding characters. It is known that the degree of this effect (referred to as an articulation coupling coefficient) is obtained by measuring the distance from a formant parameter corresponding to a character uttered alone by performing a formant analysis of the voice. The present invention utilizes the various characteristics of speech as described above to generate characters using non-verbal parameters of speech.

The method of reflecting the non-verbal voice parameter on the character can be realized, for example, by controlling the character size by the average value of the pitch frequencies of the words representing the strength of meaning. Further, since the articulatory coupling coefficient is expressed in the "gentleness" of the utterance, character data is prepared in advance along with the writing order data at the time of character generation, and the line segment data is generated in accordance with the writing order data. On the other hand, it can be expressed easily by passing through a low-pass filter, controlling the cut-off frequency of the low-pass filter by the articulation coupling coefficient, and changing the "gentleness" of the generated character.

On the other hand, as a method of inputting linguistic data, a method of converting speech into characters is known, but in the present technology, the accuracy rate of conversion is not necessarily high. Therefore, it is also effective to input language data using a keyboard and extract the non-language data from voice.

(Example) The Example of this invention is described in detail with reference to drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention. In the apparatus of this figure, the first switch of the present application is realized by switching the switch 5.
The inventions of 1 and 2 can be selectively implemented. First, an embodiment of the first invention of the present application, in which both linguistic data and non-linguistic data are extracted from speech, will be described. For initial setting, the initial setting data is sent from the keyboard 2 to the switch 5, and the setting of the switch 5 is performed. First, a voice waveform is input to the feature extraction circuit 2 via the voice waveform input terminal 1. The feature extraction circuit 2 extracts feature parameter value strings such as formant data, amplitude data, pitch data, etc. from the voice waveform and outputs them to the language data extraction circuit 3a and the non-language data extraction circuit 3b in the voice recognition unit 3, respectively. The language data extraction circuit 3a converts the characteristic parameter value string into a character string and outputs it to the standard character data memory 6 via the switch 5. The method of converting to the character string is detailed in, for example, "Voice Recognition" by Niimi (Kyoritsu Shuppan), so the description thereof is omitted here. On the other hand, the non-verbal data extraction circuit 3b extracts non-verbal data such as the articulation coupling coefficient from the characteristic parameter value sequence and outputs the non-verbal data to the character transformation processing circuit 7 via the switch 5. The character modification processing circuit 7 modifies the character data output from the standard character data memory 6 according to the character string according to the non-language data, and outputs the modified character data to the printing device 8 as modified character data. The printing device 8 prints characters according to the modified character data.

In the embodiment described above, the language data is extracted from the voice, but the embodiment of the second invention of the present application in which the language data is input from the keyboard 4 is also possible by switching the switch 5. In this case, the initialization data is input from the keyboard 4 in advance to the switch 5 via the switch control data transmission path 43 to control the switch 5, and the output of the keyboard 4 is sent via the character data transmission path 41 to the standard character data memory 6 Can be done by connecting to.

(Effect of the Invention) As described above, according to the present invention, it is possible to include non-linguistic information in a generated character string, and it is possible to generate a more familiar character string.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention. 1 ... Voice waveform input terminal, 2 ... Feature extraction circuit, 3 ...
Speech recognition unit, 3a ... language data extraction circuit, 3b ... non-language data extraction circuit, 4 ... keyboard input device, 5 ... switch, 6 ... standard character data memory, 7 ... character transformation processing circuit, 8 ...... Printer, 31,41 …… Character data transmission line, 32,42 …… Non-language data transmission line, 43 …… Switch control data transmission line.

Claims (2)

[Claims] 1. A means for converting an input voice into a character string, a means for extracting a pitch frequency variation width or a degree of articulation as non-language data from the input voice, and a character generated according to the character string and the non-language data. And a character generation device. 2. A means for inputting a character string from a keyboard,
A character generation device comprising: a unit for extracting non-language data from a voice corresponding to the character string; and a unit for generating a character according to the character string and the non-language data.