JPH0228827A - Sound recognizing device - Google Patents

Abstract

PURPOSE:To improve input efficiency by displaying all syllable candidates in the position of a cursor. CONSTITUTION:A sound recognizing device 1 equips with a microphone 2, feature extracting part 4, CPU part 5, standard pattern storing part 5, matching part 7 and an input pattern storing part 8. Then, a display part 9 and a keyboard 10 are connected. For a sound signal to be collected by this microphone 2, a feature parameter to be used for the extraction of a syllable block, etc., is computed by the feature extracting part 4. In the CPU part 5, a syllable block is determined on the basis of the said parameter and a sound pattern is prepared and stored in the input pattern storing part 8. For this sound pattern and the plural sound patterns of a standard pattern storing part 6, distance computation is executed and the syllable candidate is prepared and stored to a memory 5a. This syllable candidate is displayed on the display part 9. Thus, a user can exactly grasp how recognized the sound of the user itself is.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a speech recognition device suitably used in so-called Japanese word processors, personal computers, and the like.

BACKGROUND OF THE INVENTION In word processors, personal computers, etc., when inputting by voice, unlike input from a keyboard, for example, the uttered voice is not necessarily input accurately, and therefore the recognition result of the inputted voice is The candidates are corrected using language processing functions and used as voice input. However, if the actual voice is a lower candidate in the recognition result or is not among the candidates,
For example, syllables need to be corrected using a keyboard or the like according to cursor instructions.

FIG. 6 is a diagram showing an example of display of syllable candidates in a typical prior art. For example, when the user utters "Kokuminwo", the first candidate of the recognition results, "Bofuninwo", is first displayed as shown in FIG. 6(1). This recognition result is incorrect, and therefore the cursor indicated by reference mark A is
When the correct syllable candidates are displayed as shown in FIG. 6 (3) by using the J key etc. to move as shown in FIG. 6 (2) and operating the "↓" key, for example, When the input of the first syllable is completed, use the →J key, for example, to move cursor A to the incorrect input location after the second syllable, and such correction input operation is performed again.
As shown in 4), after all the correction input operations for the input syllables, ie, the finalization operations for the input syllables, for example, kana-kanji conversion is performed.

Problems to be Solved by the Invention In the prior art as described above, if the recognition result of the input speech is not in the top syllable candidate, the cursor is moved to that position using the "→" key etc. as described above. Then, use the [↓J key etc. to display syllable candidates sequentially from the top. Therefore, the user is unable to check the recognition results of the voice uttered from the top candidates to the bottom candidates, and therefore, if the syllable uttered by the user is not among the candidates, unnecessary key operations are required. This results in a decrease in input efficiency.

The purpose of the present invention is to simultaneously display all syllable candidates corresponding to the cursor position as the cursor moves, thereby eliminating unnecessary key operations as described above and informing the user of the internal state of the recognition result. It is an object of the present invention to provide a speech recognition device equipped with a display method that provides a clear and accurate display.

Means for Solving the Problems The present invention provides a speech recognition device that recognizes input speech syllable by syllable by calculating the distance from a standard pattern registered in advance. syllable candidate creation means for creating a syllable candidate; and display means for visually displaying a syllable in response to an output from the syllable candidate creation means, a cursor is displayed on the display means, and a cursor is displayed on the display means, and a cursor is displayed on the display means when the cursor is moved. The speech recognition device is characterized in that the syllable candidates in a syllable are displayed all at once.

Effect According to the present invention, when a voice is input to the voice recognition device,
The syllable candidate creation means creates one or more syllable candidates for each input syllable based on the recognition result. The output from the syllable candidate creation means is given to the surface water means, and of the created syllable candidates, for example, the first syllable candidate is visually displayed on the display means. A cursor is also displayed on this display means, and all syllable candidates are displayed for the syllable to which the cursor is moved.

Therefore, users can accurately grasp how their own voice has been recognized, and if the syllable they uttered is not among the candidates, there is no need to perform unnecessary key operations, and the user can quickly Correct input can be performed and input efficiency can be improved.

Embodiment FIG. 1 is a block diagram showing the configuration of a speech recognition device 1 according to an embodiment of the present invention. The audio signal picked up by the microphone 2 passes only the audio band and is provided to the feature extraction unit 4 via the amplification unit 3 which performs amplification. Feature extraction unit 4
calculates feature parameters used for extracting syllable sections and calculating distances for pattern matching, which will be described later, from the input speech waveform.

The 020 section 5 determines a syllable section based on the characteristic parameters, creates a voice pattern for the syllable section, and stores it in the input pattern storage section 8. The matching section 7 calculates the distance between the input voice pattern and a plurality of voice patterns stored in the standard pattern storage section 6.

The matching results calculated in this way are sent again to the 020 unit 5, where syllable candidates are created and stored in the memory 5a within the CPU unit 5. This syllable candidate is also displayed on the display section 9.

Display section 9 and keyboard 1 for instructing cursor movement
0 is an external device, and a device to which the speech recognition device 1 is connected, such as a word processor, is used. The CPU section 5 controls the movement of the cursor in response to input from the keyboard 10 and the like.

In the speech recognition device 1 configured as described above, when the user utters "Kokumin wo" as shown in FIG. Based on the comparison result between the stored input voice pattern and the standard pattern stored in the standard pattern storage section 6, the second
The recognition results for each syllable candidate as shown in Figure (2) are stored.

When the phrase "kokumin wo" is input, the first syllable candidate is displayed on the screen of the display unit 9, as shown in FIG. 3, and the last syllable is displayed. At the position next to the candidate, a cursor indicating the next possible input position is displayed, as indicated by reference numeral 11.

As is clear from Figure 3 above, the recognition results for the first to third syllables are incorrect, so the user must move the cursor to correct the characters they wish to input. Use the “←” key on keyboard 1o to move the cursor to
Move to syllable. As a result, on the screen of the display section 9,
As shown by reference numeral 12 in FIG. 4(1), a cursor is displayed and all syllable candidates for the first syllable where the cursor is located are displayed.

In this state, the correct character may be selected by operating the "↑" or "↓" key on the keyboard 10, and in this case, the selected syllable candidate may be replaced in the position of the first candidate, or Alternatively, the syllable candidate may be displayed separately from other syllable candidates by, for example, blinking or displaying the background color and displayed character color inverted.

When the modification of the first syllable is completed in this way, the keyboard 10
By operating the "→" key, the cursor is moved to the position of the second syllable, as indicated by reference numeral 13 in FIG. 4(2). The correction operation is performed on the second syllable in the same manner as on the first syllable, and by operating the "→" key on the keyboard 10, the cursor is moved to the position indicated by the reference mark 14 in FIG. 4(3). Thereafter, the cursor is sequentially moved as indicated by reference numeral 15 in FIG. 4 (4), and the input characters are corrected for each syllable.

When all input characters have been corrected and the cursor is moved to the input position of the next character as shown by reference mark 16 in Figure 4 (5), kana-kanji conversion, etc. is performed, and the voice is changed again. Input is made.

FIG. 5 is a flowchart for explaining the operation. When the user speaks in step n1, step ri
In step 2, a comparison with the standard pattern as described above is performed, and the recognition result of each syllable candidate is stored in the memory 5a in the CPU section 5 based on the comparison result. At step n3, the recognition result at step n2 is displayed on the display section 9.

In step n4, the "←" key or "→" key on the keyboard 10
It is determined whether a key operation has been performed, that is, whether a syllable candidate has been modified, and if so, the cursor is moved by the key operation in step n5. In step n6, all of the syllable candidates at the cursor position are displayed, and in step n7, the syllable candidates are corrected by selection of these syllable candidates, manual keystrokes, etc., and the process returns to step n3.

If the syllable candidates are not modified in step n4, that is, if the first candidate of each syllable candidate matches the input speech, processing such as kana-kanji conversion is performed in step n8, and the process returns to step r11.

In this manner, in the present speech recognition device 1, when the cursor is located in a phrase, all syllable candidates at the cursor position are displayed.

Effects of the Invention As described above, according to the present invention, all syllable candidates at the cursor position are displayed, so the user can understand how his/her own speech is recognized. can. Furthermore, in cases where the uttered syllable is not among the candidates, correction input can be quickly performed without the need for unnecessary key operations, and input efficiency can be improved.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a speech recognition device 1 according to an embodiment of the present invention, and FIG. 2 shows an example of the voice input from the microphone 2 and the recognition result of the voice in the memory 5a in the CPU section 5. FIG. 3 is a diagram showing an example of the display on the display unit 9 at the time when the voice input of one syllable is completed, FIG. 4 is a diagram showing an example of the display when each syllable candidate is selected by moving the cursor, and FIG. is a flowchart to explain the operation,
FIG. 6 is a diagram showing an example of display of syllable candidates in the prior art. 1... Voice recognition device, 2... Microphone, 3...
- Amplification section, 4... Feature extraction section, 5... CPU section, 6
...Standard pattern storage section, 7...Matching section, 8
...Input cover pattern storage section, 9...Display section, 10...
・Keyboard, 11 to 16... Cursor agent Patent attorney Number of strokes Keiichi Gokumin is set as 1 sound P registration

Claims (1)

[Claims] A syllable candidate that creates one or more syllable candidates for each syllable in a speech recognition device that recognizes input speech syllable by syllable using distance calculation from a standard pattern registered in advance. and a display means for visually displaying a syllable in response to an output from the syllable candidate creation means, wherein a cursor is displayed on the display means, and the syllable candidate for the syllable to which the cursor is moved is provided. A voice recognition device characterized by displaying all at once.