JPS6211732B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice input method, and more particularly, to a method for recognizing and displaying an input voice, and for inputting a plurality of voices having a probability of corresponding to the input voice when inputting it into an input data storage unit. Display the characters in a given priority order,
This invention relates to a voice input method that automatically selects and stores a predetermined character at the highest priority position.

Currently, data input devices include typewriters, keyboards, perforated machines, touch input devices, etc., but in order to master these devices, operators require considerable training, and the Even when performing these operations, if the operation is continued for a long time, the level of fatigue increases.

There is usually a great deal of effort involved in converting the data to be entered into a written or punched format that can be recognized by a data processing device using an input device such as the one described above.

Therefore, it would be very convenient if the data to be input into the data processing device could be directly input in voice format without the above conversion.

Research on such speech input devices has achieved considerable results, and several devices have already been realized that input data that is vocally separated and spoken words.

However, in conventional devices of this type, when the sound characteristics of the words to be input are similar, it is difficult to distinguish between them, and the recognition rate decreases. I couldn't say it was practical. Especially in cases where almost all phoneme combinations exist, such as in kana characters, it is extremely difficult to achieve near 100% speech recognition.

Regarding this problem, the patent applicant for the invention previously proposed in Japanese Patent Application Laid-Open No. 53-77402, for example,
We proposed a method for displaying several character candidates and selecting the correct one from among them.

The present invention takes the invention of the above-mentioned application one step further, and when displaying a plurality of characters that are likely to correspond to recognized speech, prioritizes the characters, automatically selects only the characters in the highest priority position, and stores the data. The purpose is to provide a voice input method that is stored in the computer.

The voice input method according to the present invention includes: an analysis extraction means for analyzing voice input and extracting features; a dictionary section storing a large amount of information for recognizing the voice input; In the voice input method, the voice input method includes means for recognizing the voice input based on the information of the voice input, and a display means for displaying a plurality of recognized candidates corresponding to the voice input from the recognition means, at a display position of the display means. The system is characterized in that it is given priority and when the candidate displayed at the highest priority display position is the desired one, it is automatically selected and stored by pronouncing the next voice input. .

Next, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 shows the configuration of an embodiment of the voice input method of the present invention, and FIG. 2 shows an operational flowchart of FIG. 1.

In FIG. 1, 1 is a microphone that provides voice data input to be recognized by the user, 2 is an operation section, and this operation section has operation keys K ₁ , K ₂ , K ₃ ,
_K4 is provided, and the character candidates (candidates...including characters) displayed on the display section, which will be described later, can be automatically or manually selected. 3 is a dictionary section, which contains a large amount of information corresponding to the input voice,
For example, "A, I, U, E, O, Ka, Ki, Ku...
……N, ga, gi, gu, ge, go, the, ji,……
"Pa, pi, .......po" are remembered. 4
5 is a frequency analysis unit that analyzes the audio spectrum of input audio, and 5 is a feature extraction unit that extracts the characteristics of the audio. Reference numeral 6 denotes a comparison/judgment unit which combines the audio information from the feature extraction unit 5 and the dictionary unit 3.
A comparative judgment is made by comparing the audio information from the source. 7 is a control section, and a frequency analysis section 4
It also gives commands to the feature extraction section 5 and controls the comparison and judgment operation of the comparison and judgment section 6. Reference numeral 8 denotes a display section, which displays the output from the comparison/judgment section 6 based on commands from the control section 7. Reference numeral 9 denotes an input data storage section for temporarily storing data (letters, numbers) corresponding to recognized correct voice input.

Next, the operation of the voice input method shown in FIG. 1 will be explained with reference to the operation flowchart shown in FIG.

As an example, consider a case where the word "Kawasaki" is pronounced into the microphone 1 and the speech is recognized. First, the user pronounces "ka". Then, the audio signal from the microphone 1 is given to the frequency analyzer 4, where the spectrum of the generated audio is electrically analyzed. This analyzed output signal is sent to the feature extraction section 5. The feature extraction section 5 extracts the features of the voice of "ka" based on the analysis results from the frequency analysis section 4, and supplies the resulting output signal to one input of the comparison/judgment section 6. As mentioned above, this comparison/determination section 6 performs the following operations:
Since it is connected to the dictionary section 3 that stores information corresponding to sounds such as clear sounds, voiced sounds, and semi-voiced sounds such as "pa, pi......", this information is extracted one after another and the feature extraction is performed. The information from section 5 is sequentially compared. Each time information that may be input voice is found, it is displayed one after another at display positions D ₁ , D ₂ , D ₃ , and D ₄ on the display unit 8 via the control unit 7 . In the illustrated embodiment, four character candidates that have a possibility (certainty) of being the input voice are displayed, for example, "ka", "ta", "pa", and "ha".

In this case, the display positions D ₁ , D ₂ ,
Among D ₃ and D ₄ , the highest priority is given to display position D ₁ , and when the character displayed at this position is the correct answer, the user pronounces the "wa" of the next word. . As a result, when the next word "wa" has been uttered and word candidates for it are sent from the comparison/judgment section 6, the control section 7 changes the character displayed at the display position _D1 to the first word. This is automatically stored in the input data storage section 9 as the first input data. Then, the word candidates sent in response to the above utterance of "wa" are displayed in the order of their likelihood.
Display on D ₁ , D ₂ , D ₃ , D ₄ .

However, the character displayed in display position D ₁ , where the character with the highest probability is supposed to be displayed, is not the correct answer, and D ₂ to D ₄ with other priorities
〓〓〓〓
If the correct answer is displayed in one of the characters, press any of the keys K ₂ , K ₃ , and K ₄ corresponding to the position where the correct character is displayed to store it in the input data storage section 9. be able to.

Similarly, the same operation is performed for "sa" and "ki" to recognize the sounds.

Also, if a correctly recognized character is not displayed in any of D ₁ to _{D 4} , press the cancel key K ₀ to cancel the display, and repeat the above steps while pronouncing the same voice again until the character is correctly recognized. Repeat.

To summarize the above recognition process using the flowchart in Figure 2, after the device starts operating in step (1),
In step (2), one character is uttered into the microphone 1 (for example, "ka"), and as a result of the above recognition operation, it is determined in step (2) whether or not the character is displayed at the display position _D1 of the display unit 8. If “YES”, step
Jump to (5) and input the next audio. If "NO", operate the selection key in step (4) and repeat the above process.

If all input voices are correctly recognized, the path of "YES" is followed in step (5), and the operation ends in step (6).

As described above, in the present invention, when the correct answer is displayed at the display position given the highest priority, the next word is uttered to automatically select it as the correctly recognized character. The data can be automatically temporarily stored in the data storage unit. If the correct answer is displayed at a different display position, it can be manually selected using the key means and stored in the data storage section.

Therefore, when the correct answer is displayed at the highest priority display position and automatically selected or performed, the user can simply pronounce it one after the other, which significantly reduces input operations and speeds up voice input. and recognition speed can be greatly increased.

[Brief explanation of the drawing]

FIG. 1 shows the configuration of an embodiment of the present invention, and FIG. 2 shows an operation flowchart of the configuration of the embodiment shown in FIG. In the figure, 1 is a microphone, 2 is an operation section, 3 is a dictionary section, 4 is a frequency analysis section, 5 is a feature extraction section, 6
7 represents a comparison/judgment section, 7 represents a control section, 8 represents a display section, and 9 represents an input data storage section. 〓〓〓〓

Claims (1)

[Claims] 1. An analysis extraction means for analyzing voice input and extracting features; a dictionary section storing a large amount of information for recognizing the voice input; A voice input method comprising a recognition means for recognizing an input, and a display means for displaying a plurality of recognized candidates corresponding to the voice input from the recognition means,
Priority is given to the display position of the display means, and when the candidate displayed at the highest priority display position is the desired one, it is automatically selected by uttering the next voice input. A voice input method characterized by being memorized.