JPH09288494A - Voice recognition device and voice recognizing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy and the processing speed of a voice recognition. SOLUTION: Recognition objective words made to be objects of the voice recognition are stored in a voice recognition data ROM 15. Moreover, words constituting sentences made to be displayed on an LCD 17 are stored in a sentence data ROM 14 by being related to corresponding recognition objective words. Then, when a voice is inputted to a microphone 1 in a state in which a certain sentence is displayed on the LCD 17, only voices of words related to words constituting the sentence being displayed on the LCD 17 among recognition objective words stored in the voice recognition data ROM 15 are recognized as objects in a voice recognizing circuit 5.

Description

Detailed Description of the Invention

[0001]

The present invention relates to a voice recognition device and a voice recognition method. In particular, the present invention relates to a voice recognition device and a voice recognition method capable of improving the voice recognition accuracy and the recognition processing speed.

[0002]

2. Description of the Related Art In a conventional electronic dictionary device, for example, when an English word is input by operating a keyboard, a phonetic symbol of the English word, information explaining the meaning (sense), and the English word are input. Example sentences used (hereinafter, as appropriate
Such information is referred to as commentary information), which is retrieved and displayed from the built-in electronic dictionary.

Further, when it is desired to obtain commentary information of an English word (for example, an English word used in an example sentence) in the commentary information retrieved in this way, the English word is also operated by the keyboard. By inputting or specifying a desired English word in the displayed commentary information by operating the cursor keys, the commentary information of the English word is searched and displayed. .

[0004]

However, it is troublesome to operate a keyboard or a cursor key to input an English word.

Therefore, all of the English words registered in the electronic dictionary built in the electronic dictionary device are set as the target of speech recognition (recognition target word), so that the English words can be input by voice. There is a way.

However, some tens of thousands of English words are registered in a practical electronic dictionary to some extent, and voice recognition of the input voice is performed for such a huge number of English words. , The recognition accuracy will deteriorate, and
The recognition processing speed will also decrease.

[0007] Further, for example, the English words used in the example sentences in the commentary information may change, but even the headings of such changed English words are registered in the electronic dictionary. In many cases, it is difficult to recognize a changed English word by voice. In addition, from the viewpoint of improving the recognition accuracy and the recognition processing speed, it may be preferable not to use words such as be verbs and prepositions that are expected to be searched infrequently as words to be recognized.

The present invention has been made in view of such a situation, and it is possible to improve the accuracy of voice recognition and the speed of recognition processing.

[0009]

According to a first aspect of the present invention, there is provided a voice recognition device in which a display word, which is a word to be displayed on a display means for displaying information, is stored in association with a corresponding recognition target word. A voice recognition unit that includes a word storage unit and that performs voice recognition on a recognition target word stored in the recognition target word storage unit displays a voice among the recognition target words stored in the recognition target word storage unit. It is characterized in that the voice recognition is performed only on the words associated with the display word displayed on the means.

According to a fourth aspect of the present invention, in the voice recognition method, a display word is displayed on the display means, and the voice recognition means selects a voice from among the recognition target words stored in the recognition target word storage means.
It is characterized in that only the words associated with the display word displayed on the display means are subjected to voice recognition.

In the voice recognition apparatus according to the first aspect, the display word storage means stores the display word, which is a phrase displayed on the display means for displaying information, in association with the corresponding recognition target word. The voice recognition unit recognizes the voice only for the recognition target words stored in the recognition target word storage unit that are associated with the display word displayed on the display unit. Has been done.

In the voice recognition method according to the present invention, the display word is displayed on the display means, and the voice is displayed on the display means among the recognition target words stored in the recognition target word storage means. The speech recognition is performed only for the words associated with the displayed word.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the configuration of an embodiment of an electronic dictionary device to which the present invention is applied. It should be noted that this electronic dictionary device is, for example, of a portable type so as to be convenient to carry, and is capable of searching English words by voice.

The microphone 1 (input means) converts the voice input therein into a voice signal which is an electric signal and outputs the voice signal to the A / D converter 2. A / D converter 2
Is adapted to sample an analog voice signal from the microphone 1 in accordance with a predetermined sampling clock and further quantize it to obtain a digital voice signal. The voice signal converted into a digital signal by the A / D converter 2 is supplied to the voice recognition circuit 5.

A RAM (Random Access Memory) 3 is
A voice signal supplied from the / D converter 2 via the voice recognition circuit 5 and data necessary for the operation of the voice recognition circuit 5 are temporarily stored. ROM (Read Only
Memory) 4 stores, for example, an application program for causing the voice recognition circuit 5 to perform voice recognition. The voice recognition circuit 5 (voice recognition means) processes the voice signal supplied from the A / D converter 2 in accordance with the application program stored in the ROM 4, so that the voice input to the microphone 1 is processed. Is voice-recognized for a recognition target word stored in a voice recognition data ROM 15 to be described later, and the voice recognition result is obtained by the CPU 1
It is designed to supply 0.

That is, the voice recognition circuit 5 is a CPU which will be described later.
According to the signal supplied from 10, some of the recognition target words stored in the voice recognition data ROM 15 are read, and a dictionary composed of the recognition target words (hereinafter, appropriately referred to as a recognition dictionary) is created. When the voice recognition circuit 5 receives the voice signal from the A / D converter 2,
It is supplied to RAM3 for storage, and then RAM
When a voice signal for one word is stored in 3, for example, the pitch (frequency), the strength (amplitude),
The speed (speech speed) is acoustically analyzed, and the likelihood of each recognition target word stored in the recognition dictionary with respect to the voice (word) input to the microphone 1 is calculated based on the analysis result. This likelihood is supplied from the voice recognition circuit 5 to the CPU 10 together with each corresponding recognition target word.

The operation unit 6 is, for example, a word designation key 6A,
It is composed of a next candidate key 6B, an enter key 6C, a scroll key 6D, etc., and is operated when giving a predetermined instruction to the device. That is, the word designation key 6A is operated when the LCD 17 described later distinguishes and displays a word that is a recognition target word and a word that is not a recognition target word. The next candidate key 6B is operated when requesting the next candidate of the voice recognition result by the voice recognition circuit 5. That is, L
As the voice recognition result by the voice recognition circuit 5, the word with the highest likelihood is first displayed on the CD 17, but the voice recognition result may be incorrect. In such a case, when the next candidate key 6B is operated, the word having the next highest likelihood is displayed on the LCD 17. The enter key 6C is operated to confirm the voice recognition result when the voice recognition result displayed on the LCD 17 is correct. Scroll key 6D
Is operated when scrolling the display of the LCD 17.

The key input circuit 9 includes a word designation key 6A, a next candidate key 6B, an enter key 6C, or a scroll key 6.
When any one of D is operated, an operation signal corresponding to the operation is output to the CPU 10. The CPU (Central Processor Unit) 10 (search means) is operated according to the operation signal from the key input circuit 9.
The voice recognition circuit 5 and the character display circuit 13 are controlled. Further, the CPU 10 is also configured to retrieve commentary information of a sentence or a word (dictionary word) stored in a sentence data ROM 14 or a word dictionary data ROM 16 described later, respectively.

The ROM 11 stores a system program and an application program for performing a predetermined process, and the CPU 11 executes various programs by executing the program stored in the ROM 11. ing. The RAM 12 stores data necessary for the operation of the CPU 10.

When the character display circuit 13 receives information such as words and commentary information from the CPU 10, the character display circuit 13 generates a bit pattern of characters constituting the information, and the LCD 1
It is designed to be supplied to the No. 7 and displayed. Other,
The character display circuit 13, under the control of the CPU 10,
The display of the LCD 17 is controlled.

The text data ROM 14 (display word storage means), the voice recognition data ROM 15 (recognition target word storage means), or the word dictionary data ROM 16 (commentary information storage means) are used for the text data, voice recognition data, or word dictionary described later. The data is stored respectively.

The LCD (liquid crystal display) 17 (display means) displays information under the control of the character display circuit 13.

Next, referring to FIG. 2, the text data ROM
14, the sentence recognition data, the voice recognition data, and the word dictionary data stored in each of the voice recognition data ROM 15 and the word dictionary data ROM 16 will be described.

First, FIG. 2A shows a sentence data ROM 1
4 shows the text data stored therein. The text data is a unique text number attached to various texts (English texts in this embodiment) to be displayed on the LCD 17 (hence, the text numbers and the texts have a one-to-one correspondence), The English words that make up each sentence (since sentences are displayed by displaying the sequence of these words, the words that make up this sentence are hereinafter referred to as display words as appropriate), and for each display word It is composed of the attached voice recognition word numbers.

In the embodiment of FIG. 2A, the display word "Seein
The sentence number 100 is attached to the sentence “Seeing is believing.” composed of “g”, “is”, and “believing”, and each display word “Seeing”, “is”, or “be” is added.
The speech recognition word number 203, 0, or 222 is attached to "lieving", respectively. In addition, the display word "Yo
The sentence "You" consisting of "u", "shall", and "see"
The sentence number 101 is attached to "should see.", and the voice recognition word number 0, 230, or 200 is attached to each display word "You", "shall", or "see". .

FIG. 2B shows the voice recognition data ROM 15
3 shows the voice recognition data stored in FIG. The voice recognition data includes a unique voice recognition word number assigned to each recognition target word (hence, there is a one-to-one correspondence between the voice recognition word number and the recognition target word), and the word as the recognition target word. (Here, it is an English word), and a dictionary word number given to each recognition target word.

In the embodiment of FIG. 2 (B), in addition to the original English words "see" and "believe",
The English words "saw", "seen", "seeing", "believed", "believing", etc., which have changed their meaning, are also recognized words. It is stored in the voice recognition data ROM 15.

Here, in this embodiment, the voice recognition word number given to the display word in FIG. 2A is the recognition target word corresponding to the display word. That is, the display word is stored in the text data ROM 14 in association with the corresponding recognition target word. Specifically, for example, the display word “Seeing” is associated with the voice recognition word number 203, and thus is associated with the recognition target word “seeing” with the voice recognition word number 203. Has been. Also, for example, in the display word "believing",
The voice recognition word number 222 is associated, and thus is associated with the recognition target word “believing” with the voice recognition word number 222. Furthermore, for example, the display word “see” is associated with the voice recognition word number 200, and thus is associated with the recognition target word “see” with the voice recognition word number 200. . As described above, the display word is linked to the recognition target word.

In this embodiment, as the voice recognition word number, for example, an integer of 0 or more is used. However, as the voice recognition word number given to the recognition target word, for example, one excluding 0, that is, a positive integer is used. Therefore, when 0 is added to the display word as the voice recognition word number, there is no recognition target word associated with the display word, and as a result, the display word is not the target of voice recognition. Not done. FIG. 2 (A)
In this example, the display words "is" and "Yo
The voice recognition word number 0 is added to "u", so that "is" and "You" are not recognized by voice.

FIG. 2C shows the word dictionary data ROM 16
The word dictionary data stored in FIG. The word dictionary data is, for example, a unique dictionary word number assigned to each English word (hereinafter, appropriately referred to as a dictionary word) published in a normal English-Japanese dictionary (hence, the dictionary word number and the dictionary word Correspond to one to one), dictionary words, and commentary information of each dictionary word.

Further, in this embodiment, the commentary information of each dictionary word is, for example, as a phonetic symbol of the dictionary word, dictionary contents including a part of speech, a variation, a meaning, and an example sentence (example) of the dictionary word. And the sentence number attached to the sentence.

In this embodiment, the dictionary word number given to the recognition target word in FIG. 2B is that of the dictionary word corresponding to the recognition target word. That is, the recognition target word is stored in the voice recognition data ROM 15 in association with the corresponding dictionary word (hence, in association with the commentary information of the dictionary word). This means that from the recognition target word to the dictionary word,
It can be said that the link has been made.

In the embodiment shown in FIG. 2, in addition to the recognition target word "see" for speech recognition of the original form of a word as a dictionary word, its variation "saw (sa)
w) ”,“ seen ”, and“ seeing ”(see
ing) ”is also associated. Furthermore, the original form of a word as a dictionary word, "bilieve", is still the recognition target word "believ" for speech recognition of the original form.
e) ”, as well as its variants“ believed ”and“ believing ”.

Further, in this embodiment, the sentence number in the commentary information of the dictionary word in FIG. 2C is attached to the sentence using the display word corresponding to the dictionary word. That is, the word dictionary data ROM 16 stores the word dictionary in association with the corresponding sentence. This means that a dictionary word is linked to a sentence.

As described above, links are established between the sentence data and the voice recognition data, between the voice recognition data and the word dictionary data, and between the word dictionary data and the sentence data.

Next, the operation will be described with reference to the flowchart of FIG. 3 and FIG. First, in step S1, the CPU 10 determines that the text data ROM
A display word forming a sentence having a predetermined sentence number and a voice recognition word number attached to the display word are read from 14, and the display word among them is displayed on the LCD 17 via the character display circuit 13.
And display it. That is, in step S1, the sentence with a predetermined sentence number is displayed on the LCD 2.

Here, the text display in step S1 is performed as follows, for example. That is, the user
When a predetermined English word is uttered, the voice is supplied to the voice recognition circuit 5 via the microphone 1 and the A / D converter 2. In this case, the voice recognition circuit 5 performs voice recognition on all the recognition target words stored in the voice recognition data ROM 15, for example, and outputs the result of the voice recognition to the CPU 10.
Output to When the CPU 10 receives an English word as a voice recognition result from the voice recognition circuit 5, the word dictionary data R
The English word is searched from the OM 16 and the sentence number attached to it is read. Further, the CPU 10 retrieves the sentence number from the sentence data ROM 14 and retrieves the display word forming the sentence with the sentence number as the sentence data R.
It is read from the OM 14 and via the character display circuit 13,
It is supplied to the LCD 17. As described above, step S
In 1, for example, a sentence (example sentence) using English words spoken by the user is displayed.

For example, when the sentence number read from the word dictionary data ROM 16 is 100, etc., in step S1, as shown in FIG. 4 (A), the sentence with the document number 100 is added (see FIG. 2 (A)). ), The display word “Se
“Seeing is believing.”, which is a sentence composed of “eing”, “is”, and “believing”, is displayed on the LCD 17.

When the sentence is displayed in step S1, the process proceeds to step S2 and the user designates the word designation key 6
The CPU 10 determines whether A has been operated. When it is determined in step S2 that the word designation key 6A is not operated, the process returns to step S2.
Further, when it is determined in step S2 that the word designation key 6A is operated, that is, the user operates the word designation key 6A, and an operation signal corresponding to the operation is
When supplied from the key input circuit 9 to the CPU 10, the process proceeds to step S3, and among the display words forming the sentence displayed in step S1 on the LCD 17, some are words to be recognized and some are not. Are displayed separately.

That is, the CPU 10 determines that the voice recognition word number read from the text data ROM 14 in step S1 is 0.
The character display circuit 13 is controlled so that, for example, an underline is added to a display word that is not defined, that is, a display word that is a recognition target word. In response to this, the character display circuit 13
Controls the LCD 17 to cause the underline to be displayed on the display words constituting the sentence displayed in step S1 whose voice recognition word number is not 0.

Here, for example, if the sentence "Seeing is believing." Is displayed as shown in FIG. 4 (A), the display word "Seeing" constituting this sentence,
As shown in FIG. 2A, the voice recognition word numbers assigned to “is” and “believing” are 203,
It is 0 or 222. Therefore, in this case, in step S3, as shown in FIG. 4B, the display words “Seeing” and “believing” whose voice recognition word number is not 0 are underlined among the display words forming the sentence. Is displayed. Therefore, in step S3, among the display words displayed on the LCD 17, only the words related to the recognition target word are underlined.

After that, the process proceeds to step S4, and the CPU 10
Supplies the voice recognition word number read from the text data ROM 14 in step S1 to the voice recognition circuit 5. When the voice recognition circuit 5 receives the voice recognition word number from the CPU 10, the voice recognition circuit 5 sets the voice recognition word number to the voice recognition data R.
A recognition dictionary is constructed by the words to be recognized, which are searched from the OM 15 and are given the voice recognition word numbers. That is, the voice recognition circuit 5 searches for a recognition target word using the voice recognition word number received from the CPU 10 as a search key, and forms a recognition dictionary with the searched recognition target word.

Here, the voice recognition word number attached to the recognition target word stored in the voice recognition data ROM 15 is
As described above, since it is a positive integer, the voice recognition circuit 5
When the voice recognition word number 0 is received from the CPU 10, the 0 voice recognition word number is ignored.

Therefore, in step S1, for example,
As shown in FIG. 4A, the sentence “Seeing is believin
g. ”is displayed, the display word“ Se
As shown in FIG. 2 (A), the voice recognition word numbers assigned to “eing”, “is”, and “believing” are as follows.
Since it is 203, 0, or 222, the speech recognition circuit 5 ignores 0 of these, and recognizes the remaining 203 or 222 as a speech recognition word number, respectively, “Seeing” or “Seeing”. Believing (be
lieving) ”forms a recognition dictionary.

Regarding steps S3 and S4, the process of step S4 is performed first, and then step S3.
It is also possible to perform the processing of step 3, or step S
It is also possible to perform the processes of 3 and S4 at the same time.

As described above, after the display word which is the recognition target word is underlined and the recognition dictionary is constructed from the recognition target words corresponding to the display word, the user:
When it is desired to obtain commentary information on an underlined display word (English word), the English word is uttered. The voice uttered by the user is converted into an analog voice signal by passing through the microphone 1, and is further converted into a digital voice signal by passing through the A / D converter 2. This digital voice signal is passed through the voice recognition circuit 5 in step S5.
It is supplied to the RAM 3 to be stored (acquired). When the voice signal is stored in the RAM 3, the voice recognition circuit 5 proceeds to step S6 and determines whether or not the voice signal for one word is stored in the RAM 3. When it is determined in step S6 that the voice signal for one word is not yet stored in the RAM 3, the process returns to step S5, whereby the voice signal is continuously stored in the RAM 3.

When it is determined in step S6 that the voice signal for one word is stored in the RAM3, the process proceeds to step S7, and the voice recognition circuit 5 determines in step S4 based on the voice signal stored in the RAM3. Targeting only the recognition target words registered in the configured recognition dictionary,
Recognize the voice input to the microphone 1. That is, the voice recognition circuit 5 calculates the likelihood of each of the recognition target words stored in the recognition dictionary with respect to the voice (word) input to the microphone 1, and associates it with the voice recognition word number of each recognition target word. And supply it to the CPU 10.

Therefore, in the voice recognition circuit 5, step S
7, among the recognition target words stored in the voice recognition data ROM 15, only the words related to the display word displayed on the LCD 17 are subjected to the voice recognition, that is, the number of words is small. Since voice recognition is performed for words, the voice recognition data ROM 1
It is possible to improve the recognition accuracy and the recognition processing speed as compared with the case where the voice recognition is performed on all the recognition target words stored in FIG.

Further, in step S3, the words to be recognized by the speech recognition circuit 5 are underlined so that they are displayed separately from the words not to be recognized by speech. Therefore, the user can easily recognize the word that is currently the target of voice recognition (or the word that is not the target of voice recognition).

When the CPU 10 receives the voice recognition word number and the likelihood of each recognition target word registered in the recognition dictionary from the voice recognition circuit 5, in step S8, the highest likelihood of the voice recognition word numbers. Select the one associated with. Further, the CPU 10 is a display word associated with the recognition target word corresponding to the selected voice recognition word number (hereinafter, appropriately referred to as the selected voice recognition word number), which is displayed on the LCD 17. The character display circuit 13 is controlled so that it can be recognized as the first candidate of the voice recognition result, for example, in reverse display.

Thus, for example, as described with reference to FIG. 4B, the display words "Seeing" and "believin" are displayed.
When "g" is the recognition target word, for example, "S
When the likelihood is high in the order of “eeing” and “believing”, in step S8, “Seeing” with the highest likelihood is highlighted as shown in FIG. 4C.

After that, the process proceeds to step S9, and the enter key 6
The CPU 10 determines whether C has been operated. When it is determined in step S9 that the enter key 6C has not been operated, the process proceeds to step S10, in which the CPU 10 determines whether or not the next candidate key 6B has been operated. In step S10, the next candidate key 6B
If it is determined that is not operated, the process returns to step S9, and step S9 is repeated until it is determined in step 9 that the decision key 6C has been operated, or until it is determined in step S10 that the next candidate key 6B has been operated. And the processing of S10 is repeated.

If it is determined in step S10 that the next candidate key 6B has been operated, that is, the LCD 1
In order that the display word highlighted in 7 is not the word spoken by the user but the word with the next highest likelihood,
If the user operates the next candidate key 6B, step S
In step 11, the CPU 10 newly selects a voice recognition word number associated with the next highest likelihood, and recognizes the voice recognition word number corresponding to the newly selected voice recognition word number (selected voice recognition word number). The character display circuit 13 is controlled so that the display word associated with the target word, which is displayed on the LCD 17, is highlighted. As a result, the display word having the next highest likelihood is highlighted in place of the display word currently highlighted.

That is, for example, as described with reference to FIG. 4C, in the case where "Seeing" having the highest likelihood of the display words "Seeing" or "believing" is highlighted, When the next candidate key 6B is operated,
As shown in FIG. 4D, instead of “Seeing”, “believing” having the next highest likelihood is highlighted.

After that, the process returns to step S9, and the processes after step S9 are repeated again. Then, step S9
In the case where it is determined that the enter key 6C is operated, that is, the display word highlighted on the LCD 17 is the word uttered by the user, and the user needs to confirm the word as the final voice recognition result. When the enter key 6C is operated, the process proceeds to step S12, and the CPU 10 searches the word dictionary data ROM 16 for the commentary information of the display word.

That is, the CPU 10 causes the dictionary word number (FIG. 2) associated with the voice recognition word number which is the selected voice recognition word number when the enter key 6C is operated.
(B)) is recognized by referring to the voice recognition data ROM 15 via the voice recognition circuit 5. Furthermore, CPU
10 searches the word dictionary data ROM 16 for the recognized dictionary word number.

Then, the CPU 10 proceeds to step S13, controls the character display circuit 13 so as to display the commentary information associated with the retrieved word dictionary number, and ends the processing. As described above, the search is performed by using the word dictionary number as the search key, whereby the LCD 17 displays the commentary information corresponding to the dictionary word associated with the confirmed voice recognition result.

That is, for example, when the display word "Seeing" is highlighted as shown in FIG. 4C, when the enter key 6C is operated, the display word "Seeing" is associated with the display word "Seeing". Dictionary word number 30 associated with the recognized word “Seeing”
The commentary information of the dictionary word "see" (FIG. 2 (C)) of 0 is retrieved and displayed. Further, for example, as shown in FIG. 4D, when the display word "believing" is highlighted, when the enter key 6C is operated, it is associated with the display word "believing". The dictionary word “believe” of the dictionary word number 302 associated with the recognition target word “believing” (FIG. 2
The explanation information of (C)) is retrieved and displayed.

As described above, the voice recognition data ROM 1
5, the original English words "see" and "believe (b)
"elieve)", its variations "saw", "seen", and "seeing",
"Believed" and "believed"
Since "lieving)" and the like are also registered as recognition target words, it is possible to recognize not only the original form but also the modified form by voice.

Further, the original form "see" and its variations "saw", "seen", and "seeing" are all original dictionary words "." It is associated with "see". Similarly,
"Believe", "believe"
d) ”, and“ believing ”are also associated with the original dictionary word“ believe ”. Therefore, the commentary information of the original form can be obtained from the word dictionary data in which the commentary information of only the original form is registered, regardless of whether the utterance of the original form or the variation form is made. In other words, it is not necessary to prepare commentary information for each original form or variation, and the data amount of the word dictionary data is smaller than that in the case where separate commentary information is prepared, which is, so to speak, efficiency. It becomes possible to compose typical word dictionary data.

Further, since the display word and the recognition target word are associated with each other and the recognition target word and the dictionary word are also associated with each other, the display words having the same notation but having different pronunciations and meanings ( For example, even though the past tense and the past molecular tense are the same notation but have different pronunciations, or homonyms, each displayed word is related to the correct dictionary word via the recognition target word. As a result, the wrong dictionary word is not searched even though the recognition result of the voice is correct. Specifically, for example, in a certain sentence A, the display word "scrap" meaning a fragment or a small piece is converted into a dictionary word "sc
rap ”, and the display word“ scrap ”that means conflict or quarrel in the other sentence B is translated into the dictionary word“ sc ”that means conflict or quarrel through the recognition target word.
When the sentence A or B is displayed and the display word “scrap” is uttered, the word dictionary data ROM 16 indicates that the dictionary word of the correct meaning is That is, the dictionary word “scrap” meaning a fragment or a small piece or the dictionary word “scrap” meaning a fight or a quarrel are respectively searched.

Further, by making the display word stored in the text data ROM 14 the target of voice recognition and the target of searching commentary information, the voice recognition data RO
In the M15 and the word dictionary data ROM 16, for example,
Data about proper nouns can also be included. That is, it is difficult to set proper nouns as targets of speech recognition and retrieval of commentary information without any restrictions, because of the number of proper nouns in the world. Data ROM among the proper nouns
If only the words stored in 14 are displayed,
The number thereof is not so large, and therefore, it can be a target of voice recognition and a target of searching commentary information.

In step S13, the dictionary contents (pronunciation symbols and meanings of dictionary words) of the commentary information are
What is stored in the word dictionary data ROM 16 is displayed as it is, but for example sentences, the word dictionary data R is displayed.
As shown in FIG. 2C, since the sentence number corresponding to the example sentence is registered in the OM 16, the sentence (display word string) corresponding to the sentence number is stored in the sentence data ROM 14
It is designed to be searched and displayed from.

Therefore, in this embodiment, when a certain word in the sentence is uttered while the sentence is displayed, the sentence data ROM 1 is used as an example sentence using the word.
Since the sentence as the display word string stored in No. 4 is displayed, the commentary information of the words forming the sentence can be searched again.

Next, the process of forming the recognition dictionary in step S4 of FIG. 3 will be described in more detail with reference to FIG.

In the embodiment of FIG. 5A, for example, two sentences "Second thoughts are best." And "Seeing is"
“Believing.” is retrieved from the sentence data ROM 14 and the entire two sentences are displayed on the LCD 17. In this example, the be verb "ar
Display words "Second" and "thought" excluding "e" and "is"
"s", "best", "Seeing", and "believing" are underlined, so in this case, step S4
Then, a recognition dictionary is composed of these five words, and is used as the target of the voice recognition performed by the voice recognition circuit 5.

On the other hand, in the embodiment of FIG. 5B, for example,
Four sentences "Second thoughts are best.", "Seeing i
s believing., '' Slow but steady wins the rac
e. ", and" So many countries, so many costom
"s." is retrieved from the text data ROM 14 and the LCD 1
7, the "Seeing is believing." And "Slow but s
"teady wins the race." is displayed. However, for the sentence "Slow but steady wins the race."
Only "ow but steady wins the" is displayed, but "race." is not displayed.

In this embodiment, the be verb "i"
Display words "Seeing", excluding "s" and the article "the"
lieving, "Slow", "but", "steady", and "wins" are underlined, so in this case,
In step S4, a recognition dictionary is constructed from these 6 words and is the target of the voice recognition performed by the voice recognition circuit 5.

That is, in this case, the display word "race" not displayed on the LCD 17 and the sentence "Second thoughts"
are best. "and" So many countries, so many cost
The display words that make up "oms." are not uttered by the user. Therefore, since such a display word does not need to be the target of voice recognition, only the display word actually displayed on the LCD 17 (however, the be verb and the article are excluded here) is targeted. Then, a recognition dictionary is constructed.

Then, the user operates the scroll key 6
D is operated, which causes the text "Second tho" to be displayed on the LCD 17, for example, as shown by the dotted line in FIG.
If "ughts are best." and "Seeing is believing." are displayed, the display words that make up these sentences (however, here, the be verb is excluded) "Seco
nd, thoughts, best, Seeing, and be
"Lieving" is underlined, and a recognition dictionary is constructed for these.

The display word currently displayed on the LCD 17 is adapted to be recognized by the CPU 10. The voice recognition circuit 5 creates a recognition dictionary based on the recognition result of the display word by the CPU 10. It is designed to be done.

However, the recognition dictionary is constructed not only for the display words actually displayed on the LCD 17 but also for the display words constituting the sentence retrieved from the sentence data ROM 14. It is also possible.

Although the present invention has been described in the case of being applied to an electronic dictionary device, the present invention can be applied to any device that inputs a word or phrase displayed on the display means for displaying information by voice. Applicable.

In this embodiment, step S1 in FIG.
In the above, the word is searched and the sentence which is an example sentence using the word is displayed. However, in addition, for example, the sentence data ROM 14 stores a sentence such as a book or a newspaper, and the power of the device is turned on. When it is turned on, it is possible to display the text of such a book or newspaper in step S1.

In the present embodiment, the text data ROM 1
4, the voice recognition data ROM 15 and the word dictionary data ROM 16 are incorporated in the device. However, the sentence data ROM 14, the voice recognition data ROM 15, and the word dictionary data ROM 16 may be, for example, an IC card that is removable from the device. It is possible to make it built in.

Further, in this embodiment, the electronic dictionary device
Although the function of a so-called English-Japanese dictionary has been provided, the electronic dictionary device can be provided with the functions of, for example, a Japanese-English dictionary and dictionaries for other languages in addition to the English-Japanese dictionary. is there.

Further, in the present embodiment, the word is set as the target of the voice recognition, but it is also possible to set the word as the target of the voice recognition in addition to the word.

Furthermore, in the present embodiment, it is determined whether or not the display word is the target of voice recognition, by the sentence data ROM 15
The voice recognition word number stored in (FIG. 2A) is 0.
Whether or not the display word is the target of voice recognition is stored in the text data ROM 15 in addition to this, for example, information indicating that fact (hereinafter, appropriately referred to as determination information). It is also possible to add it to each displayed word and make a discrimination based on this discrimination information. Then, in this case, it is possible to set a specific display word as a target of voice recognition or not to perform it, depending on a predetermined condition. That is, for example, as the discrimination information, one of 0, 1, and
In addition to adding to the display word, a predetermined button is provided, and when the predetermined button is not operated,
Only display words with discrimination information of 0 are targeted for voice recognition, and when a predetermined button is operated, display words with discrimination information of either 0 or 1 are targeted for voice recognition. You can do so.

Further, in this embodiment, the text data ROM 1
Although the English sentence (display word string) is stored in 4, the sentence data ROM 14 stores the Japanese translation in addition to the English sentence, for example. Things are possible.

Further, in this embodiment, the text data ROM
In FIG. 14, one speech recognition word number is associated with one display word. However, for example, a display word that is expected to have a plurality of patterns of utterance If there is a display word), such plural patterns of utterance (this utterance may be erroneous as the pronunciation of the display word) are stored in the voice recognition data ROM 15 as a recognition target word. It is possible to associate the voice recognition target numbers of the plurality of recognition target words with the plurality of utterance display words. That is, it is possible to associate a plurality of recognition target words with one display word. Specifically, for example, the display word is "ISO (International Organization f
or Standardization) ”, the words to be recognized are“ ISO (ISO) ”,“ ISO (ISO) ”,
It is also possible to associate "ISO" and "YES". In this case, "iso",
By associating "iso" and "yes" with the commentary information of "ISO", no matter which of these is uttered, the utterance will be treated as corresponding to the display word "ISO". It becomes possible to recognize and retrieve the commentary information.

In the present embodiment, the voice recognition circuit 5 supplies the CPU 10 with the voice recognition word numbers and the likelihoods of the respective recognition target words corresponding thereto, but the voice recognition circuit 5 supplies the CPU 10 with the likelihood. For example, it is possible to supply only the speech recognition word numbers in descending order of likelihood of the respective recognition target words corresponding thereto. In this case, the CPU 10 may determine the recognition target word having a high likelihood from the arrangement of the voice recognition word numbers.

Furthermore, in this embodiment, the LCD 17
Although there is a plurality of words associated with the same voice recognition word number (here, excluding 0) among the displayed words in the sentence displayed in, no particular mention was made. In such a case, the CPU 10
The character display circuit 13 is controlled so that all such display words are underlined, and one of the plurality of voice recognition word numbers is used for voice recognition as a component of the recognition dictionary. It is adapted to transmit to the circuit 5.

Further, in the present embodiment, the recognition target word corresponding to the original form of the verb and the recognition target word corresponding to its variation are related to the dictionary word corresponding to the original form. For example, all recognition target words corresponding to nouns having the same meaning can be related to a dictionary word corresponding to any one of such nouns.

[0084]

According to the voice recognition device of the first aspect and the voice recognition method of the fourth aspect, in the state in which the display word is displayed, the voice is one of the predetermined recognition target words. Only the words associated with the displayed word are displayed for voice recognition. Therefore, since the voice recognition is performed using the minimum necessary recognition target words that the user may utter, it is possible to improve the voice recognition accuracy and the voice recognition processing speed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of an electronic dictionary device to which the present invention is applied.

FIG. 2 is a diagram illustrating sentence data, voice recognition data, or word book data stored in a sentence data ROM 14, a voice recognition data ROM 15, or a word dictionary data ROM 16 of FIG. 1, respectively.

3 is a flowchart for explaining the operation of the electronic dictionary device of FIG.

FIG. 4 is a diagram showing a display state of LCD 17.

FIG. 5 is a diagram for explaining the process of step S4 of FIG.

[Explanation of symbols]

1 microphone (input means), 5 voice recognition circuit (voice recognition means), 10 CPU (search means), 14 text data ROM (display word storage means), 15 voice recognition data ROM (recognition target word storage means), 16 words Dictionary data ROM (commentary information storage means), 17 LCD
(Display means)

Claims (4)

[Claims] 1. A recognition target word storage unit that stores a recognition target word that is a phrase that is a target of voice recognition, an input unit that inputs a voice, and a voice that is input to the input unit. A voice recognition device comprising voice recognition means for recognizing the recognition target word stored in the word storage means, the display means displaying information, and the display word being a phrase displayed on the display means. Further comprising a display word storage unit that stores the associated target recognition word in association with the recognition target word, wherein the voice recognition unit stores the voice in the recognition target word stored in the recognition target word storage unit. Of the above, the voice recognition device is characterized in that the voice recognition is performed only on the object associated with the display word displayed on the display means. 2. The voice according to claim 1, wherein the display unit distinguishes and displays the display words that are associated with the recognition target word from the display words. Recognition device. 3. The comment information storage means for storing commentary information for explaining a phrase, and the search means for searching the commentary information from the commentary information storage means, wherein the recognition target word storage means are: The recognition target word is stored in association with the corresponding commentary information, and the search unit searches for the commentary information associated with the recognition target word that has been speech-recognized by the voice recognition unit. The voice recognition device according to claim 1, wherein the display unit displays the commentary information retrieved by the retrieval unit. 4. A recognition target word storage unit that stores a recognition target word that is a phrase that is a target of voice recognition, an input unit that inputs a voice, and a voice that is input to the input unit as the recognition target. A voice recognition means for recognizing the recognition target word stored in the word storage means as a target, a display means for displaying information, and a display word which is a phrase displayed on the display means, corresponding to the recognition target word. A voice recognition method for a voice recognition device, comprising: a display word storage unit that stores the display word, the display unit displaying the display word, and the voice recognition unit recognizes the voice. A speech recognition method characterized in that among the recognition target words stored in the target word storage means, only the words associated with the display word displayed on the display means are subjected to voice recognition. Law.