JPH0764593A - Information processor and control method - Google Patents

Abstract

PURPOSE:To provide an information processor and control method in which a key at the time of extracting only a necessary part can be obtained by reflecting a pronouncing state peculiar to a person who inputs a voice on the display of speech information, and only the necessary part can be automatically extracted. CONSTITUTION:This device is equipped with a microphone 19 which inputs the speech information, display part 16 which displays the speech information, and memory 22 for preservation which stores the speech information. A CPU 11 detects the pronouncing state of the interval, up/down, and strength or the like of the pronunciation of each syllable of the speech information inputted through the microphone 19, and the speech information is displayed on the display part 16 in a display configuration in which the pronouncing state can be recognized at the time of displaying the voice information. Also, the CPU 11 detects a key word, topic switching word, or related word from among the voice information, extracts the speech information according to the detected result, and stores it in the memory 22 for preservation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information processing apparatus capable of inputting voice information, and more particularly to an information processing apparatus and a control method for displaying input voice information and extracting only a necessary portion.

[0002]

2. Description of the Related Art Conventionally, there is an information processing apparatus for performing voice processing, in which voice information input through a microphone is displayed on a display device so as to be shown to a user. This is realized by voice-recognizing the input voice information and performing text conversion into a corresponding character code.

By the way, when inputting voice information,
You don't always enter in a constant tone. That is, for example, considering that a sentence such as "Kyouwa Tenki is very good" is input by voice, you can enter it with some gap between "Kyoha" and "Tenki" or "Very".
You can pronounce it strongly in the part of, or raise it with the ending "is" and enter it. This varies depending on the voice input person.

Conventionally, in this type of information processing apparatus, the text-converted voice information is simply listed and displayed, and the pronunciation interval for each syllable, the intonation such as rising / falling, and the strength of the sound. It was not possible to reflect the pronunciation condition such as the accent in the display.

When inputting voice information, it is rare to input only necessary information from the beginning to the end.
It also contains informative information. Therefore, it is necessary to extract only the necessary portion from the display screen later, but in the conventional model, the user himself / herself did this extraction.

[0006]

As described above, conventionally, since the input voice information is simply displayed in a list, the pronunciation state peculiar to the voice input person cannot be judged from the display, which is necessary later. There was a problem that it was difficult to grasp the clue when extracting only the part from the display screen.

Further, conventionally, when extracting a necessary part,
Since the user always had to do it himself, it was very troublesome, and there was a problem that it was difficult to extract it especially when there was a lot of information (when the talk was long).

The present invention has been made in view of the above points, and can be used as a clue when extracting only a necessary portion by reflecting the pronunciation state peculiar to the voice input person in the display of voice information. It is another object of the present invention to provide an information processing device and a control method capable of automatically extracting only a necessary part.

[0009]

The present invention comprises an input means for inputting voice information, a display means for displaying the voice information, and a storage means for storing the voice information. The voice information input by the above is recognized by voice, and text conversion is performed into character code information.
Detecting pronunciation intervals such as pronunciation intervals, rising / falling, and strength for each syllable, and displaying the text-converted voice information on the display means in a display form in which the pronunciation status can be understood. Is.

Further, the voice information within the designated range is extracted from the displayed voice information, and the extracted voice information is stored in the storage means.

The present invention further comprises input means for inputting voice information and storage means for storing the voice information, wherein the voice information input by the input means is voice-recognized and character code information is obtained. Text conversion is performed, and a preset keyword, word for topic conversion or a word related to the keyword is detected from the voice information converted into text, and the word for topic conversion or the related word is detected. When the keyword is detected or when the related word is detected more than a specified value, the voice information is extracted from the detected position, and the extracted voice is stored. The information is stored in the storage means.

[0012]

With the above arrangement, when voice information is input,
For each syllable, a pronunciation state such as a pronunciation interval, rising / falling, and strength is detected. And when displaying the audio information,
The display form is such that the pronunciation state can be understood. Therefore, when instructing the extraction range on the display screen, it is possible to easily extract only the necessary portion by using the sound generation state as a clue.

Further, according to the above configuration, the keyword, the topic conversion word or the related word is detected from the voice information. Then, when the topic conversion word or the related word is detected, the detected position is held as the extraction position of the voice information, and when the keyword is detected or when the related word is detected more than the specified value, from that position Audio information is extracted. Therefore, it is possible to extract the voice information of only the necessary portion without requiring the extraction operation by the user.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a main part of an information processing apparatus capable of voice input. In FIG. 1, the CPU 11
Is for controlling the entire device.
2 and the RAM 13 are accessed, and a voice input control process or the like is executed by activating a program according to the input instruction.

The ROM 12 has a program area 12a for storing a program required for starting the CPU 11.
A font area 12b for storing font data used for display, a dictionary area 12c for storing a voice dictionary required for voice recognition, and neural network information for connecting related words. The network area 12d is provided. The above-mentioned voice dictionary is a dictionary in which voice waveforms and hiragana characters are associated with each other, and also has patterns such as whispers and consonants.

The RAM 13 has a table area 13 for storing a display table for displaying voice information.
a, a voice buffer 1 for temporarily storing voice information
3b, a keyword area 13c for storing keywords, an extraction position storage area 13d for storing extraction positions, and the like are provided. In the display table, as shown in FIG. 2, a character code which is a recognition result of voice information, a timer cout indicating a pronunciation interval, rising / falling information indicating pronunciation intonation, and a strength indicating a pronunciation accent. Information is stored for each syllable.

The input unit 14 is composed of, for example, a keyboard and is used to input information such as character codes and commands. The input control unit 15 controls the input of the information input by the input unit 14. The display unit 16 is composed of, for example, a liquid crystal display device (LCD), and displays various kinds of information including audio information. The display control unit 17 uses a VRAM (video RAM) 18 to perform display control on the display unit 16. Display data to be displayed on the display unit 16 is stored in the VRAM 18.

The microphone 19 is for inputting voice information. The voice control unit 20 controls the input of voice information input through the microphone 19. The timer 21 is C
The time is counted according to the instruction of PU11. The storage memory 22 stores sound information for storage.

Next, the operation of the embodiment will be described. Here, (a) voice information input operation, (b) voice information display operation, (c) voice information extraction operation by user instruction,
(D) The operation for automatically extracting the voice information by the system will be described separately.

(A) Input operation of voice information In the present invention, when voice information is input through the microphone 19, the voice information is voice-recognized and the result is displayed on the display unit 16. In this case, it is simply displayed. Instead of this, it is characterized in that the pronunciation intervals of each syllable, intonation such as rising and falling, and pronunciation states such as accents that are the strength of the sound are reflected in the display.

As shown in the flow chart of FIG. 3, when input of voice information is instructed through the input unit 14 (step A1), the CPU 11 activates the timer 21 (step A2) and waits for input of voice information (step A2). N in step A3
o).

The voice information is input through the microphone 19, processed by the voice controller 20 and then given to the CPU 11. When the CPU 11 inputs the voice information that has been voice-processed by the voice control unit 20 (Y in step A3).
es), one syllable is cut out, and the current time Tn is detected from the timer 21 (step A4).
An initial value "0" is set as a count value in the display table provided in the table area 13a of No. 3 (step A5).

Further, the CPU 11 uses the voice dictionary provided in the dictionary area 12c of the ROM 12 to perform voice recognition for the one syllable (step A6), and the character code (Hiragana character corresponding to the voice information) as the recognition result. ) Is set on the display table (step A7). Further, the CPU 11 detects the rising / falling information and the strength information of the one syllable, and sets each level as the initial value "1" in the display table (steps A8 and A).
9).

When the next voice information is input (Yes in step A10), the CPU 11 cuts out the next section and the current time Tn + from the timer 21 in the same manner as above.
1 is detected (step A11), and Tn + 1-Tn (sound generation interval) is set as a count value in the display table (step A12). Further, the CPU 11 performs voice recognition for the one syllable (step A13), and sets the character code (Hiragana character corresponding to voice information) as the recognition result in the display table (step A1).
4). Further, the CPU 11 detects the rising / falling information and the strength information of the one syllable, and sets the comparison value of each level when the initial level is "1" in the display table (steps A15 and A16).

In this way, the same processing as described above is repeated until the input of voice information is completed (No in step A10). As a result, in the display table, as shown in FIG. 2, the recognition result (Hiragana characters), the pronunciation interval, the up / down information, and the strength information are set for each voice.

(B) Display operation of voice information When the input of voice information is completed, the CPU 11 executes the display process shown in the flowchart of FIG. This display processing is performed according to the display table created as described above.

That is, the CPU 11 first reads out the character code of the first syllable set in the display table and displays the corresponding character pattern on the display unit 16 (step B1). Here, if the next character exists (Yes in step B2), the CPU 11 displays the character. At that time, the count value corresponding to the character code set in the display table is read and the count value is read. A space corresponding to the value is displayed (step B3). Then, the CPU 11 adds the up / down added to the character code.
The next character pattern is displayed on the display unit 16 in a display form corresponding to the falling information and the strength information (step B4).

FIG. 5 shows a display example. For example, when a sentence such as “Today is very good” or “I am a sacred person ...” is input by voice, the result of voice recognition for each syllable is displayed on the display unit 16. Displayed on the display screen. In this case, a space corresponding to the interval is displayed between the characters so that the interval of pronunciation can be known for each syllable.

Further, as the display position of "ne" in the figure is shifted upward, the intonation is pronounced higher or lower than the sound of the initial "ki" as a reference. The characters are displayed by shifting up and down according to the level at that time.

As shown in the enlarged view of "very" in the figure, a character whose accent is stronger or weaker than the sound of the first "ki" is used as a reference. The image is enlarged or reduced in size according to the current level.

(C) Extraction Operation of Voice Information According to User's Instruction Next, an operation of extracting only a necessary portion from the voice information displayed on the display screen will be described.

As shown in the flowchart of FIG. 6, the user designates the start point and the end point of the voice information to be extracted with a cursor or the like (steps C1 and C2). CPU1
1 sets the start point and the end point as an extraction range, and extracts voice information within the extraction range (step C3). Specifically, the display table stored in the table area 13a of the RAM 13 is referred to, and the character code existing in the extraction range is read from the table. Then, the CPU 11 stores the voice information thus extracted as the voice information for storage in the storage memory 22 (step C4).

This state is shown in FIG. In the display as shown in Fig. 5, "from Japanese" to "is"
When the voice information up to is required, the cursor 31 is set to "ni" which is the beginning of the context, and then the cursor 32 is set to "su" which is the end of the context. As a result, only the portion "is Japanese ..." is extracted and stored in the storage memory 22.

In this way, by designating the extraction range on the display screen, it is possible to save the voice information of only the necessary portion. In this case, since the pronunciation state peculiar to the voice input person is reflected in the display of the voice information as described in (a) above, only the necessary portion can be easily extracted with the pronunciation state as a clue.

(D) Automatic Extraction Operation of Voice Information by System In the above (c), the user himself / herself extracted the voice information. However, the feature is that the system automatically extracts the voice information.
This automatic extraction is performed based on three pieces of information of a keyword, a word for topic conversion, and a related word. The keyword has been previously specified by the user and is stored in the keyword area 13c of the RAM 13. The word for topic change is, for example, a word such as “Well” or “By the way”. A related word is a word related to a keyword,
For example, if "Japanese processing" is the keyword,
"Natural language processing", "word processor", etc. correspond to that. These related words are obtained from the neural network information stored in the network area 12d of the ROM 12.

Here, here, a voice dictionary in which a voice waveform and a word are associated with each other (a dictionary having an accuracy that can be decomposed into morphemes)
It is assumed that the input voice information is replaced with a word and displayed by using, and each word such as a keyword, a word for topic conversion, and a related word is searched from the word. The voice information replaced with words is temporarily stored in the voice buffer 13b of the RAM 13.

As shown in the flow charts of FIGS. 7 and 8, the CPU 11 first clears the word search counter A and the related word appearance counter B to "0" as initial settings, and switches the topic. The flag F is turned off and reset (step D1). The word search counter A is for counting the number of words searched from the search target. The related word appearance counter B is for counting the number of words related to the keyword. The topic conversion flag F is turned on when a topic conversion word such as “Well” or “By the way” is retrieved. The counters A and B and the flag F are provided in the RAM 13, for example.

The CPU 11 cuts out one word at the beginning of the sentence from the voice information stored in the voice buffer 13b (step D2) and updates the word search counter A (step D3). Then, the CPU 11 checks whether the extracted word is the same word as the keyword, a word for topic conversion, or a word related to the keyword (steps D4 to D6).

When the same word as the keyword is detected in step D4, the CPU 11 checks whether or not the topic conversion flag F is on (step D9). this is,
Even if a keyword is detected, we do not know whether it was used in the main subject or happened to be used outside the main subject, so avoid extracting voice information using only the keyword, and extract it only after a topic change. This is because.

When the topic conversion flag F is off (No in step D9), the CPU 11 checks whether or not the value of the related word appearance counter B has reached the specified value (step D10). This is prepared in case the main subject is entered without changing the topic.

When the value of the related word appearance counter B has not reached the specified value (No in step D10), C
The PU 11 returns to the processing from step D3, updates the word search counter A, and searches for the next word. If the value of the related word appearance counter B has reached the specified value (Yes in step D10), or if the topic conversion flag F is on in step D9, the CPU 11 keeps it in step D15 or step D18 described later. The voice information is extracted based on the extraction position (step D11) and stored in the storage memory 22 (step D12).

When the topic conversion word is detected in step D5 and the topic conversion flag F is in the off state (No in step D13), the CPU 11 turns on the flag F (step D14), After the detected position at that time is kept as the extraction position in the extraction position storage area 13d (step D15), the process returns to step D3,
The word search counter A is updated to search for the next word.

When the word related to the keyword is detected in step D6, the CPU 11 updates the related word appearance counter B (step D16). At this time, if the value of the related word appearance counter B is "1", that is, if the related word is detected for the first time (Yes in step D17).
s), the CPU 11 keeps the detection position at that time as the extraction position in the extraction position storage area 13d (step D
18) Then, returning to the processing from step D3, the word search counter A is updated to search for the next word. On the other hand, when the value of the related word appearance counter B has reached the specified value (Yes in step D19), the CPU 11 extracts voice information from the extraction position kept in the extraction position storage area 13d (step D20). The data is stored in the storage memory 22 (step D21).

In this way, words are taken out one by one from the beginning of the sentence, and it is checked whether or not it is a keyword, topic conversion word, or related word. Then, when the topic conversion word or the related word is detected, the detected position is held as the extraction position of the voice information, and when the keyword is detected or when the related word is detected more than the specified value, from that position Audio information is extracted.

When the word search counter A reaches a prescribed value (step D7), in preparation for the case where the position to be extracted cannot be searched even if a predetermined number of words are searched, the word search counter A is used. , After clearing all the related word appearance counter B and the topic conversion flag F (step D8),
The search shall be started again.

FIG. 9 shows a display example. FIG. 9A shows an example of a case where voice information is extracted using "Japanese processing" as a keyword in a sentence having a topic conversion word such as "By the way". In this case, the position of “By the way” is kept as the extraction position, and when the keyword “Japanese processing” is detected, the voice information is extracted based on the position of the “By the way”.

Further, FIG. 9B shows an example in which voice information is extracted from the related words in a sentence in which there is no topic conversion word such as "By the way" and no keyword such as "Japanese processing". In this case, for example, “image”, “natural language processing”, and “information” are searched as words related to the keyword “Japanese processing” from a neural network as shown in FIG. , The audio information is extracted based on the position of the related word detected first (the first “image”).

In the above embodiment, only the case of detecting the beginning of the context has been described, but the same method as described above can be used for detecting the end of the context. In this case, for example, a word indicating the end of the context such as "as above", "after all", or "as a concluding" is detected, or a neural network is used to find a large number of words far from the keyword. There are methods such as detecting.

[0049]

As described above, according to the present invention, pronunciation states such as intervals, rising / falling, and strength of the pronunciation of the input voice information are detected for each syllable, and when the voice information is displayed, this pronunciation is detected. Since the display mode is such that the state can be understood, when the extraction range is designated on the display screen, the pronunciation state can be used as a clue to easily extract only the necessary portion.

Further, according to the present invention, the keyword, the topic conversion word or the related word is detected from the voice information and the voice information is extracted according to the detection result. Therefore, the extraction operation by the user is required. Instead, it is possible to extract only the necessary part. With this, in the information processing apparatus capable of inputting voice information, only a necessary portion is automatically extracted from the input voice information, and the burden on the user can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an information processing apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram showing the configuration of a display table of the embodiment.

FIG. 3 is a flowchart for explaining a voice information input operation of the embodiment.

FIG. 4 is a flowchart for explaining a voice information display operation of the embodiment.

FIG. 5 is a diagram showing a display example of voice information according to the embodiment.

FIG. 6 is a flowchart for explaining a voice information extracting operation according to a user's instruction in the embodiment.

FIG. 7 is a flowchart for explaining an automatic extraction operation of voice information by the system of the embodiment.

FIG. 8 is a flowchart for explaining an automatic extraction operation of voice information by the system of the embodiment.

FIG. 9 is a diagram showing a display example of audio information according to the embodiment.

FIG. 10 is a diagram showing a configuration of a neural knit work according to the same embodiment.

[Explanation of symbols]

11 ... CPU, 12 ... ROM, 12c ... Dictionary area, 12
d ... network area, 13 ... RAM, 13a ... table area, 13b ... voice buffer, 13c ... keyword area, 13d ... extraction position storage area, 14 ... input section, 15 ...
Input control unit, 16 ... Display unit, 17 ... Display control unit, 18 ...
VRAM, 19 ... Microphone, 20 ... Voice control unit, 21 ... Timer, 22 ... Storage memory.

Continuation of the front page (51) Int.Cl. ⁶ Identification number Office reference number FI Technical display location G06F 3/16 320 D 7165-5B H 7165-5B 17/22

Claims (11)

[Claims] 1. Input means for inputting voice information, voice recognition of the voice information input by the input means, text conversion to character code information, and detection of the pronunciation interval for each syllable. An information processing apparatus comprising: a voice processing means; and a display control means for displaying the voice information converted into text by the voice processing means in a display form in which the pronunciation interval can be understood. 2. Input means for inputting voice information, voice recognition of the voice information input by the input means, text conversion into character code information, and rise / fall of the pronunciation for each syllable. An information processing apparatus comprising: a voice processing means for detecting; and a display control means for displaying the voice information converted into text by the voice processing means in a display form in which rise / fall of the pronunciation can be recognized. . 3. Input means for inputting voice information, voice recognition of the voice information input by the input means, text conversion to character code information, and detection of the strength of pronunciation for each syllable. An information processing apparatus comprising: a voice processing means; and a display control means for displaying the voice information converted into text by the voice processing means in a display form in which the strength of pronunciation is understood. 4. An instructing means for instructing a range to be extracted from the audio information displayed by the display control means, and an extracting means for extracting the audio information within the extraction range in accordance with an instruction by the instructing means. The information processing apparatus according to claim 1, further comprising a storage unit that stores the voice information extracted by the extraction unit. 5. Input means for inputting voice information, voice recognition of the voice information input by the input means, text conversion into character code information, and interval of pronunciation of each syllable, rise / fall. Voice processing means for detecting a pronunciation state such as falling or strength, display control means for displaying the voice information converted into text by the voice processing means in a display form so as to understand the pronunciation state, and the display control means. Instructing means for instructing a range to be extracted with respect to the displayed voice information, extracting means for extracting the voice information within the extraction range in accordance with an instruction by the instructing means, and the voice extracted by the extracting means. An information processing apparatus comprising: a storage unit that stores information. 6. An input unit for inputting voice information, a display unit for displaying the voice information, and a storage unit for storing the voice information, wherein the voice information input by the input unit is stored. The voice recognition is performed, the text is converted into character code information, and the pronunciation states such as the pronunciation interval, rise / fall, and strength are detected for each syllable, and the pronunciation state of the text-converted voice information can be understood. While displaying on the display means in a display form, the voice information within a designated range is extracted from the displayed voice information, and the extracted voice information is stored in the storage means. Control method characterized by 7. An input means for inputting voice information, a voice processing means for voice-recognizing the voice information input by the input means, and text-converting it to character code information, and text conversion by the voice processing means. Detecting means for detecting a preset keyword from the above-mentioned voice information, extracting means for extracting the voice information based on the keyword detected by the detecting means, and the extracting means for extracting the voice information. An information processing apparatus comprising: a storage unit that stores voice information. 8. An input means for inputting voice information, a voice processing means for voice-recognizing the voice information input by the input means, and text-converting it to character code information, and text conversion by the voice processing means. Detecting means for detecting a word for topic conversion from the above-mentioned voice information, extracting means for extracting the voice information based on the word for topic conversion detected by the detecting means, and this extracting means An information processing apparatus comprising: a storage unit that stores the extracted voice information. 9. Input means for inputting voice information, voice processing means for voice-recognizing the voice information input by the input means, and text-converting it to character code information, and text-converting by the voice processing means. Detecting means for detecting a word associated with a preset keyword from among the above-mentioned voice information, extracting means for extracting the voice information based on the related word detected by the detecting means, and this extracting means An information processing apparatus, comprising: a storage unit that stores the voice information extracted by. 10. Input means for inputting voice information, voice processing means for voice-recognizing the voice information input by the input means, and text-converting it to character code information, and text conversion by the voice processing means. Detecting means for detecting a preset keyword, word for topic conversion or a word related to the keyword from the above-mentioned voice information, and the word for topic conversion or the related word is detected by this detecting means. When the keyword is detected, or when the related word is detected more than a specified value, the extraction means for extracting the voice information from the detection position, and the extraction means An information processing apparatus comprising: a storage unit that stores the extracted voice information. 11. Input means for inputting voice information,
A storage unit for storing the voice information is provided, the voice information input by the input unit is voice-recognized, the text is converted into character code information, and preset from the text-converted voice information. A keyword, a word for topic conversion or a word related to the keyword is detected, and when the word for topic conversion or the related word is detected, the detection position is held and the keyword is detected. At this time or when the related word is detected at a specified value or more, the voice information is extracted from the detection position, and the extracted voice information is stored in the storage means.