JPS59184940A - Voice word processor - Google Patents

Abstract

PURPOSE:To simplify the operating procedure and to improve the recognition factor for a voice word processor by sorting the voice input into a single sound, a voiced sound, a semi-voiced sound and a contracted sound respectively to produce a standard pattern and then selecting a pair of voice out of said pattern memory when the voice is fed. CONSTITUTION:The voice input fed from a microphone 1 undergoes a frequency analysis through an analyzing part 2 and is delivered to an input pattern part 3. An unknown input voice pattern of the part 3 is compared 5 with one of plural standard voice patterns stored in a standard voice pattern memory part 4 and sent to a deciding part 6 for the final decision. The part 4 stores the voice input by sorting it into four groups, that is, a single sound 40, a voiced sound 41, a semi-voiced sound 42 and a contracted sound 43 respectively. Then a standard voice pattern is designated with operation of a keyboard (not shown in the diagram) and undergoes the comparison processing. The voice 40 is delivered to a comparison processing part 5 with no operation of keys. In this case, a voiced sound key 400, for example, is pushed and therefore a pattern of the semi-voiced sound pattern of a memory 41 is delivered for comparison 5. This eliminates the unnaturalness of the voice input and improves the recognition factor of an voice word processor.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice word processor that inputs monosyllables such as "i", "ro", and "ha" by vocalizing them.

In recent years, voice input has attracted attention as a new input method to replace keyboard input, and one of its applications is being researched for use in word processors, and some of it is being crystallized, but it has not yet been put into full-scale practical use. Think about the situation. The reason for this is that compared to conventional keyboard input, Bentouch input, etc., the operations are multiple and practical efficiency is not improved. The input audio is also good.

``ro'' and ``ha'' have 45 syllables with voiced sounds, semi-voiced sounds, and ↑yu sounds.

When foreign words (fi, pho), etc. are added, there are over 100 monosyllables, but a device that can recognize all of these monosyllables has not yet been put to practical use.

What are the audio word processors currently in use?

If all single syllables, including voiced sounds, semi-voiced sounds, and persistent sounds, are recognized, the recognition rate will decrease, so only the 45 syllables such as "i", "ro", and "ha", excluding voiced sounds, semi-voiced sounds, and persistent sounds, are recognized. It is said that To do this, for example, if you want to input the voiced sound "ga", say "ka" once and input it, then immediately press the keyboard with the voiced sound "ka J" to "ga" and enter "ga". input. Another example of 1flJ is to input ``ka'' by saying it once, and then immediately after uttering a control word such as ``dakuten'', ``ka'' is converted into ``J'' and ``ga'' is input. ing. In both examples, the voice word processor cannot recognize voiced sounds, so this is a method to compensate for this, but for the operator's voice, it is necessary to input "ka" to input "ga".
Since it is unnatural to say ``voice'', inputting voiced sounds etc. is a complicated operation.

SUMMARY OF THE INVENTION An object of the present invention is to provide a voice word processor that eliminates the unnaturalness of conventional speech and that is easier to operate.

In other words, according to the present invention, when inputting, for example, "ga", the voice word processor performs recognition processing specifically for the voiced sound by pressing the keyboard of the voiced sound immediately before, and the operator pronounces the voiced sound "ga" itself. According to the present invention, the utterance of ``ka'' is extremely natural compared to the conventional method of uttering ``ka''. Furthermore, when a key such as a voiced sound is pressed, the recognition result at that time is limited to those within that group, thereby preventing erroneous recognition as outside the group. With the present invention, it is possible to provide a high-performance speech word processor that can recognize all of the hundred or so monosyllables mentioned above using the current level of recognition technology.

The present invention will be described in detail below with reference to drawing f. mark,
FIG. 1 shows an embodiment according to the present invention. In this figure,
The audio signal from the microphone 1 undergoes frequency analysis in the analysis section 2 and is output to the input cover section 3. Next, the unknown input M voice pattern in the input cover pattern section 3 is compared with the standard voice pattern stored in advance in the standard voice pattern storage section 4 in the comparison processing section 5, and the result is sent to the determination section 6. In the end, any single syllable is determined and the result is output.

FIG. 2 is an enlarged view of the keyboard section of the audio word processor. A keyboard is provided that corresponds to voiced sounds, semi-voiced sounds, etc.

Signals 400 to 402 normally indicate "0" level,
The pressed key tg+ta, level "1", is output.

I! FIG. 3 shows the details of the standard voice pattern storage section 4 shown in FIG. 1. 40 is "i""ro" [is J4545 sound standard voice pattern storage part, 41 is kekuon, 42 is semi-voiced, 4
3 is a storage unit for standard sound patterns. Normally, when no key is pressed, il- is sent to signals 400 to 402 from Fig. 2.
Since all are at the "0" level, a standard speech pattern of 45 syllables is output.

Also, when a voiced sound is pressed, the signal 400 goes to the 1" level, so the standard pattern of the voiced sound is output. Similarly, the signal 4
When 01 and 402 are each "l", the standard pattern of the semi-voiced s-sulphon is output.

Although pattern matching has been described above as an embodiment of the present invention, it is a matter of course that the present invention can be applied to any other recognition method.

[Brief explanation of drawings]

FIG. 1 shows the recognition section of a spoken word processor. FIG. 2 is an enlarged view of the keyboard section of the audio word processor. FIG. 3 is a detailed diagram of the standard voice pattern storage section. In the diagram, 1...microphone, 2...
Audio signal analysis section, 3... Input cover turn section, 4.
...Standard voice pattern storage unit, 5...Comparison processing unit, 6...Discrimination unit, 40...Standard voice (45 syllables Tubaturn memory n!XH' +1...
... Standard speech (semi-voiced sound) pattern memory section. 42 ... Standard speech (semi-voiced sound) pattern memory section,
43... Ritsu "P quasi-voice (sulky sound) pattern storage unit,
400...Dark sound keyboard output signal, 401...
...Semi-voiced keyboard output signal, 402... Sound keyboard output signal.

Claims (1)

[Claims] It is characterized by comprising a plurality of standard speech patterns corresponding to monosyllables, voiced sounds, semi-voiced sounds, and persistent sounds, and recognizing one of the plurality of sound standard patterns by selecting one of the plurality of sound standard patterns and comparing it with an input speech pattern. Speech word processor.