JP2001134290A - System and method for speech recognition, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To facilitate the document preparation by speech recognition. SOLUTION: A speech inputted by a speaker with a microphone 1A is transferred to a speech recognition device 2 through an interface 1B as digital speech information. A speech recognition part 2A converts the received speech information into a text consisting of not less than one character. An intonation score allocating part 2B allocates variation in the speed, sound volume, and frequency of the speech information to each converted character as an intonation score. This intonation score is temporarily stored in a storage device 3. A score mark creating part 4A allocates display attributes such as a color attribute and a font attribute to each character in the text converted by the speech recognition part 2A on the basis of the intonation score stored in the storage device 3. A display data creating part 4D creates display data having the attribute allocated to each character in the text according to the display conditions stored in a display condition storage part 4C, and makes the display device 5 display it.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition system and method, and a computer-readable recording medium on which a program for realizing such a system is recorded. Attached and output.

[0002]

2. Description of the Related Art Conventionally, there is a voice recognition system that recognizes voice input from a microphone, converts the voice into text, and outputs the text. Japanese Patent Application Laid-Open No. H11-197991 discloses, among such systems, a system which analyzes a change in a phoneme pattern or a change in voice power in an input voice and recognizes even the emotion of a speaker. I have.

[0003]

However, the above-mentioned system, when a speaker utters one word,
It is assumed that the word is recognized by speech and the emotion level is recognized (see paragraphs 007 to 007 in the above publication).
3, 0077-0080). That is, when a speaker utters a sentence composed of a plurality of words, and outputs the text as a text composed of a continuous character string by voice recognition, it does not consider how the emotion level is reflected in the text. Therefore, even if the above-mentioned system is applied to document creation, document creation cannot always be facilitated.

[0004] It is an object of the present invention to provide a speech recognition system and method capable of facilitating document creation by speech recognition, and a computer-readable recording medium on which a program for the system is recorded.

[0005]

To achieve the above object, a speech recognition system according to a first aspect of the present invention comprises:
Voice recognition means for converting voice information input from the outside into a character string; generating change information indicating a change in voice indicated by the voice information based on the voice information; And a change information assigning means for assigning a display attribute for each predetermined unit in the character string according to the change information assigned by the change information assigning means. It is characterized by the following.

In the above-described speech recognition system, a display attribute corresponding to a change in speech is automatically added to each predetermined unit in the character string converted by the speech recognition means. For example,
At the part where the speaker puts intonation and emphasizes the voice, a display attribute for emphasizing the part can be added even in the character string. For this reason, according to the speech recognition system, it is easy to create a document by speech recognition.

In the above speech recognition system, the change information allocating means generates a score based on at least one of a speed, a volume and a frequency of the voice indicated by the voice information, and generates the generated score in the character string. Can be assigned for each predetermined unit. in this case,
The attribute adding unit may add a display attribute for each predetermined unit in the character string based on the score generated by the change information allocating unit.

Further, the attribute adding means may add a display attribute to a predetermined unit in the corresponding character string when the score generated by the change information allocating means exceeds a predetermined threshold.

The speech recognition system may further include syntax analysis means for syntax-analyzing the character string recognized by the speech recognition means. In this case, the attribute adding means includes:
Further, according to the result of parsing by the parsing means,
A display attribute may be added for each predetermined unit in the character string.

[0010] The speech recognition system may further include learning means for learning a tendency of the change information assigned by the change information assigning means to a specific one of the predetermined units in the character string. In this case, the attribute adding unit may further add a display attribute for each predetermined unit in the character string according to the learning result accumulated in the learning unit.

By providing the parsing means and the learning means as described above, a display attribute is added to an unnatural position in a character string, and a predetermined unit thereof is unnaturally generated in a finally generated character string. There is no emphasis.

In the above speech recognition system, the display attribute may be, for example, a color attribute or a font attribute.

In the above speech recognition system, the predetermined unit to which the display attribute is added is, for example, a character, word, phrase, section,
It can be a sentence or the like.

In order to achieve the above object, a voice recognition method according to a second aspect of the present invention includes a voice recognition step of converting voice information input from the outside into a character string, and the voice recognition method based on the voice information. Change information indicating a change in the voice indicated by the voice information, and the generated change information is allocated in the change information allocating step of allocating each of the predetermined units in the character string converted in the voice recognition step; An attribute adding step of adding a display attribute for each predetermined unit in the character string according to the change information.

[0015] To achieve the above object, a computer-readable recording medium according to a third aspect of the present invention comprises:
Speech recognition means for converting speech information input from the outside into a character string, generating change information indicating a change in speech indicated by the speech information based on the speech information, and generating the change information by the speech recognition means. A change information allocating means for allocating each predetermined unit in the converted character string, and a computer device as an attribute adding means for adding a display attribute for each predetermined unit in the character string according to the change information allocated by the change information allocating means. It is characterized by recording a program for functioning.

[0016]

Embodiments of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the configuration of the speech recognition system according to this embodiment. As shown in the figure, the speech recognition system includes a speech information input device 1,
The apparatus includes a voice recognition device 2, a storage device 3, a display data creation device 4, and a display device 5.

The voice information input device 1 is a user interface device for influencing a user's voice (voice), and includes a microphone 1A for converting a voice uttered by the user into an analog electrical signal, and an analog electrical signal output from the microphone 1A. Is converted into digital audio information at a predetermined sampling interval.

The voice recognition device 2 has an interface 1B
Generates a character string (text) composed of one or more characters to which an inflection score (described later) is attached for each character from the converted digital voice information.
And an intonation score allocating unit 2B.

The speech recognition section 2A converts speech information input from the interface 1B into text by applying a speech recognition technique. Various techniques conventionally known in the field of speech recognition and kana-kanji conversion can be used as the speech recognition technique applied by the speech recognition unit 2A to convert speech information into text.

The intonation score allocating unit 2B creates an intonation score indicating the degree of the time stamp (speed), volume (volume) and pitch (frequency) of the voice indicated by the voice information input from the interface 1B, and converts this into a voice. It is assigned to each character in the text converted by the recognition unit 2A.

The storage device 3 stores the intonation score assigning unit 2B.
Temporarily stores the text in which each character is assigned an intonation score.

The display data creation device 4 creates display data to be displayed on the display device 5 from the text assigned the intonation score stored in the storage device 3, and includes a score marker creation unit 4A and a display condition setting unit. 4B, a display condition storage unit 4C, and a display data creation unit 4D.

The score marker creating section 4A reads the information stored in the storage device 3, and based on the intonation score assigned to each character in the text by the intonation score assigning section 2B, assigns each character a color attribute and a font attribute. Assign display attributes.

The display condition setting section 4B sets display conditions for displaying text on the display device 5 by a user operation. The display condition storage unit 4C includes a display condition setting unit 4B.
The display conditions set from are stored. The display conditions include display attributes such as the number of lines on the display device 5, the number of characters in the lines, the default character color, font type, and size.

The display data creating section 4D stores the score marker creating section 4 according to the display conditions stored in the display condition storing section 4C.
A creates display data to be displayed on the display device 5 from the text in which the display attribute is assigned to each character.

The display device 5 is a CRT (Cathode Ray Tub).
e), a user interface device composed of an LCD (Liquid Crystal Display) or the like, displaying the display data created by the display data creation unit 4D and presenting it to the user.

This speech recognition system is, for example,
It is realized by a general-purpose computer such as a personal computer. Here, the speech recognition device 2 and the display data creation device 4 are a processor and a memory for storing its program,
It is realized by a memory for storing data, and may actually be realized on the same computer system. Programs for implementing these devices 2 and 4 may be recorded on a computer-readable recording medium such as a CD-ROM and distributed, or may be distributed via a network. The storage device 3 is realized by providing an area in an external storage device or the like of the general-purpose computer.

Hereinafter, the operation of the speech recognition system according to this embodiment will be described with reference to the flowchart of FIG.

First, the user of the system performs a predetermined operation to instruct the start of voice input (step S1).
1). Next, the user utters a voice corresponding to the text to be generated toward the microphone 1A. The voice is converted into an analog electric signal by the microphone 1A, converted into digital voice information by the interface 1B, and passed to the voice recognition unit 2A (step S12).

Next, the speech recognition section 2A converts the speech information passed from the interface 1B into a text consisting of one or more characters by applying a speech recognition technique. The voice recognition unit 2A converts the converted text into a time stamp (speed) and a volume (volume) of the voice extracted from the voice information.
And the information indicating the pitch (frequency) is passed to the intonation score assigning unit 2B (step S13).

Next, the intonation score assigning unit 2B assigns the amount of change in the time stamp, volume, and pitch to each character in the text as an intonation score. The intonation score assigning unit 2B temporarily stores the text in which the intonation score is assigned to each character in the storage device 3 (step S1).
4). The details of this process will be described later.

Next, based on the intonation score stored in the storage device 3, the score marker creating unit 4A assigns display attributes such as a color attribute and a font attribute to each character in the text converted by the speech recognition unit 2A. Score sign making unit 4
A transfers the text to which the display attribute has been assigned to the display data creation unit 4D (step S15). The details of this process will be described later.

Next, the display data creation section 4D creates display data in which display attributes are assigned to each character in the text passed from the score marker creation section 4A in accordance with the display conditions stored in the display condition storage section 4C. I do. Display data creation unit 4
D passes the created display data to the display device 5 (step S16).

Then, the display device 5 displays the display data passed from the display data creation unit 4D (step S1).
7). Then, the process of this flowchart ends.

FIG. 3 shows that in step S14 of FIG.
It is a flowchart which shows the process which the intonation score allocation part 2B performs in detail. The intonation score assignment unit 2B
According to the time stamp passed from the voice recognition unit 2A,
The change amount of the speed of the voice input by the user from the microphone 1A is calculated. The calculated change amount of the voice speed is temporarily stored as the internal data D1 (step S2).
1).

Next, the intonation score assigning unit 2B calculates the relative change of the data scalar amount of the voice input from the microphone 1A by the user according to the volume and the pitch passed from the voice recognition unit 2A. The data scalar amount calculated here is also temporarily stored as the internal data D1 (step S22).

The intonation score allocating unit 2B determines which character in the text passed from the voice recognition unit 2A is related to the variation in the speed of the voice and the relative variation in the data scalar, which are temporarily stored as the internal data D1. Judge whether it is a corresponding one or a combination with a character. Then, the intonation score assigning unit 2B assigns these combined changes to each character in the text as an intonation score (step S23). Then, the process of this flowchart ends.

FIG. 4 shows that in step S15 of FIG.
It is a flowchart which shows the process which the score marker preparation part 4A performs in detail. This processing is performed for each character in the text with the intonation score stored in the storage device 3.
Each is done. That is, the score marker creating unit 4A
Repeats this process for the number of characters.

First, the score marker creating section 4A determines whether or not the intonation score given to the character is below a preset threshold D2 (step S31). If it is determined that it is below the threshold value (step S31; Y
ES), the score indicator creating unit 4A ends the processing of this flowchart for the character.

On the other hand, if it is determined that the score is not below the threshold, that is, if there is an inflection score equal to or greater than the threshold (step S31; NO), the score marker creating section 4A determines whether or not to map a color attribute to the character. It is determined according to the intonation score to be performed (step S32). If it is determined that the color attribute should be mapped (step S32; YE
S), the score marker creating unit 4A maps a color attribute corresponding to the intonation score to the character (step S33).

If it is determined that the color attribute should not be mapped (step S32; NO), or if the mapping of the color attribute has been completed, the score marker creating unit 4A applies the font attribute (for example, font size) to the character. Is to be mapped according to the corresponding intonation score (step S34).

If it is determined that the font attribute should not be mapped (step S34; NO), the score marker creating section 4A ends the processing of this flowchart for the character. On the other hand, when it is determined that the font attribute should be mapped (step S34; YE
S), the score marker creating unit 4A maps a font attribute corresponding to the intonation score to the character (step S3).
5). Then, the processing of this flowchart for the character is completed.

Hereinafter, texts displayed on the display device 5 in the voice recognition system according to this embodiment will be described with reference to specific examples. Here, for simplicity of explanation, it is assumed that score marker creating section 4A gives only font attributes as display attributes. Further, it is assumed that the user has input two sentences, “I was very happy today” and “Oh, I was done.” From the microphone 1A.

In the first sentence, it is assumed that the user speaks while emphasizing the "very" part. Then, the intonation score of these three characters becomes larger than the other parts and exceeds the threshold. Therefore, the score indicator creating unit 4A adds a font attribute for making the font size larger than the default font size to the three characters. This allows
As shown in FIG. 5, the word “very” is displayed on the display device 5 with being emphasized more than other portions.

In the second sentence, it is assumed that the user speaks the portion of "O", for example, by gradually reducing the voice. Then, the intonation score in this part gradually decreases. Here, assuming that the voice recognition unit 2A converts this part into six characters “Ooooooo”, the score marker creating unit 4A adds a font attribute to each character so that the font size decreases. . As a result, as shown in FIG. 5, the font size is displayed on the display device 5 while the font size is decreasing, and the sound is decreasing.

As described above, in the speech recognition system according to this embodiment, the intonation score assigning unit 2B assigns an intonation score to each character in the text converted by the speech recognition unit 2A, and based on this intonation score. Thus, the score marker creating section 4A assigns display attributes such as color attributes and font attributes.

That is, when a user inputs a voice from the microphone 1A, a text with a display attribute can be automatically created. Here, the character corresponding to the portion where the intonation is emphasized in the input voice can be emphasized even in the text by changing the character color or increasing the font size. . For this reason, according to the speech recognition system according to this embodiment, it is possible to easily create a document by speech recognition.

The present invention is not limited to the above embodiment,
Various modifications and applications are possible. Hereinafter, modifications of the above-described embodiment applicable to the present invention will be described.

In the above embodiment, the intonation score assigning unit 2B assigns the intonation score in units of characters,
The score marker creating unit 4A attaches display attributes such as color information in units of characters. However, the display attribute may be added in units of words. Further, a display attribute may be added in units of a phrase or a section in a sentence or in units of a sentence.

In the above embodiment, the intonation score assigning unit 2B creates an intonation score based on the time stamp, volume, and pitch extracted from the audio information, and assigns the intonation score to each character in the text. However, the intonation score assigning unit 2B can create an intonation score using any one or more of these.
Alternatively, change information based on changes in voices other than these three may be created, and a display attribute may be added to each character in the text according to the change information.

In the above-described embodiment, the score marker creating section 4A assigns a color attribute and a font attribute as display attributes to each character in the text. However, other display attributes, such as bold, italic,
Character decoration attributes such as underline, upper line, shading, background color, superscript, subscript, conversion to upper case (when the character is an alphabet), inversion and rotation of the character may be assigned to each character. Also, not only the font size but also the font type (Gothic, Mincho, etc.) may be assigned as the font attribute.

In the above embodiment, the score marker creating section 4A assigns a color attribute to each character in the text converted by the speech recognition section 2A based on only the intonation score assigned to each character by the intonation score assigning section 2B. Alternatively, display attributes such as font attributes are assigned. However, the score indicator creating unit 4A may assign a display attribute to each character based on other information.

FIG. 6 is a block diagram showing a speech recognition system according to another embodiment. The speech recognition system of FIG. 6 has a morphological analysis unit 6 in addition to the configuration of FIG.
Having. The morphological analysis unit 6 morphologically analyzes the text generated by the speech recognition unit 2A by converting the speech information, and passes the result to the score marker creation unit 4A. The morphological analysis means 6 may be realized on the same computer as the speech recognition device 2 and / or the display data creation device 4.

In this case, the score marker creating section 4A
Can further assign a display attribute to each character in the text based on the analysis result of the morphological analysis means 6. By grammatically analyzing the text generated by the speech recognition unit 2A in this way, it is possible to prevent the display device 5 from displaying an unnatural text in which a part of speech that is not usually emphasized, such as a particle or an auxiliary verb, is emphasized. can do.

FIG. 7 is a block diagram showing a voice recognition system according to still another embodiment. The speech recognition system of FIG. 7 further includes a learning unit 7 in addition to the configuration of FIG. The learning means 7 learns the tendency of score assignment by the intonation score assigning section 2B for the specified specific word, and reports the learning result to the score marker creating section 4A.
Pass to.

In this case, the score marker creating section 4A
Can further assign a display attribute to each character in the text based on the learning result of the learning means 7. By learning the tendency of the intonation score assigned by the intonation score assigning unit 2B in this manner, for example, words that are always emphasized due to the accent of the user's words and the like are prevented from being emphasized in the text, Unnatural text can be prevented from being displayed on the display device 5.

Further, the above-described speech recognition system may further include control means for controlling whether or not the score marker creating section 4A attaches a display attribute to each character in the text. In this case, it is assumed that the user can give an instruction as to whether or not to add a display attribute to each character in the text to the control means, and the control means controls the score marker creating unit 4A based on the instruction. Is also good.

In the above embodiment, the voice information input device 1 including the microphone 1A and the interface 1B is applied as the user interface device for inputting voice information to the voice recognition unit 2A, but the present invention is not limited to this. I can't. For example, the voice previously recorded on a cassette tape or the like may be converted into digital voice information and passed to the voice recognition unit 2A. Alternatively, voice information may be read from a medium in which voice information in which voice has been sampled in advance is recorded and input to the voice recognition unit 2A.

In the above embodiment, the display device 5 is applied as the user interface device for outputting the generated text to the outside, but the present invention is not limited to this. For example, a printer that prints and outputs text on a medium such as paper may be used. Further, the communication device may be a communication device that transmits the data to a remote terminal device via a network and outputs the data from the terminal device.

[0061]

As described above, according to the present invention,
Since a display attribute corresponding to a change in the voice is automatically added to the character string converted from the voice information, it is easy to create a document by voice recognition.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a speech recognition system according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a process executed by the voice recognition system of FIG. 1;

FIG. 3 is a flowchart showing in detail a process executed by an intonation score assigning unit.

FIG. 4 is a flowchart showing in detail a process executed by a score marker creating unit.

FIG. 5 is a diagram showing an example of a text displayed on a display device in the voice recognition system of FIG. 1;

FIG. 6 is a block diagram showing a configuration of a speech recognition system according to another embodiment of the present invention.

FIG. 7 is a block diagram showing a configuration of a speech recognition system according to another embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 1 voice information input device 1A microphone 1B interface 2 voice recognition device 2A voice recognition unit 2B intonation score allocation unit 3 storage device 4 display data generation device 4A score marker generation unit 4B display condition setting unit 4C display condition storage unit 4D display data generation unit 5 display device 6 morphological analysis means 7 learning means

Claims (9)

[Claims] 1. A speech recognition means for converting speech information input from the outside into a character string, and generating change information indicating a change in speech indicated by the speech information based on the speech information, and generating the generated change information. Change information assigning means for assigning a display attribute to each predetermined unit in the character string according to the change information assigned by the change information assigning means. A speech recognition system comprising an adding unit. 2. The change information allocating means generates a score based on at least one of speed, volume and frequency of a voice indicated by the voice information, and allocates the generated score to each predetermined unit in the character string. 2. The speech recognition system according to claim 1, wherein the attribute adding unit adds a display attribute for each predetermined unit in the character string based on the score generated by the change information allocating unit. 3. The attribute adding means, when the score generated by the change information allocating means exceeds a predetermined threshold,
3. The speech recognition system according to claim 2, wherein a display attribute is added to a predetermined unit in the corresponding character string. 4. The apparatus according to claim 1, further comprising: a parsing means for parsing the character string recognized by said voice recognition means, wherein said attribute adding means further comprises: for each predetermined unit in said character string in accordance with a result of parsing by said parsing means. The speech recognition system according to any one of claims 1 to 3, wherein a display attribute is added to the information. 5. The apparatus according to claim 1, further comprising: a learning unit configured to learn a tendency of the change information allocated by the change information allocating unit to a specific one of the predetermined units in the character string. The speech recognition system according to any one of claims 1 to 4, wherein a display attribute is added for each predetermined unit in the character string according to a learning result accumulated in a learning unit. 6. The speech recognition system according to claim 1, wherein the display attribute is a color attribute. 7. The speech recognition system according to claim 1, wherein said display attribute is a font attribute. 8. A voice recognition step of converting voice information input from the outside into a character string, and generating change information indicating a change in voice indicated by the voice information based on the voice information, and generating the generated change information. And a change information assigning step of assigning a display attribute for each predetermined unit in the character string according to the change information assigned in the change information assigning step. And a speech recognition method. 9. Voice recognition means for converting voice information input from the outside into a character string, generating change information indicating a change in voice indicated by the voice information based on the voice information, and generating the generated change information. Change information allocating means for allocating each predetermined unit in a character string converted by the voice recognition means, and attribute addition for adding a display attribute for each predetermined unit in the character string in accordance with the change information allocated by the change information allocating means A computer-readable recording medium on which a program for causing a computer device to function as means is recorded.