JP3741536B2 - Educational equipment - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an educational device using voice recognition.
[0002]
[Prior art]
In recent years, several technologies have been proposed as language education equipment using speech. For example, in Japanese Patent Laid-Open No. 63-303400, a model voice (a model voice of a language (for example, English) that you want to learn) is put in a card-like recording device, and it is repeated while listening to it to practice speaking. Techniques for performing are shown. According to this technology, the model voice can be effectively heard, but there is a drawback that even if there is a mistake in pronunciation, it cannot be corrected unless the person himself / herself notices.
[0003]
On the other hand, in Japanese Patent Laid-Open No. 59-220775, an exemplary voice is recorded on a magnetic tape, and the voice of a user is compared with that to determine mechanically whether they are similar or not and talk about the results. The technology to inform the person is shown. Japanese Patent Application Laid-Open No. 60-162281 stores the voice of the model speaker and the input voice of the practitioner, and analyzes them to compare the characteristics of the input voice of the practitioner with the characteristics of the model voice. Evaluation, the analysis and the evaluation result are displayed on the display device, the practitioner hears his / her pronunciation, the analysis result of the model voice and his / her voice displayed on the display device 4, and the evaluation result of his / her voice Shows a technique for checking the difference in voice characteristics between the model voice and one's own voice, and correcting one's pronunciation.
[0004]
[Problems to be solved by the invention]
According to the techniques of Japanese Patent Laid-Open Nos. 59-220775 and 60-162281, the model voice and the learner's voice are compared with each other in amplitude, pitch, and formant. For humans, there is a drawback that it is difficult to understand what should be corrected to get close to the model voice.
[0005]
Further, in the techniques disclosed in Japanese Patent Laid-Open Nos. 59-220775 and 60-162281, only a model voice (a model voice of a language to be learned (for example, English)) is output. However, even if you can acquire the correct pronunciation of this language, you have the disadvantage that you cannot know it immediately when you want to know what the language means. In other words, language learning is based on the acquisition of words. In order to acquire words with the correct pronunciation, it is very important not only to show the correct pronunciation as a teaching material but also to understand the meaning. In Japan, a technique is widely used in which Japanese is written on the front of the word card and foreign language is written on the back, and the words in the foreign language (for example, English) are reminded by looking at the Japanese on the front. However, the techniques disclosed in Japanese Patent Laid-Open Nos. 59-220775 and 60-162281 have a drawback that they cannot be used like the above word cards.
[0006]
The present invention presents a model voice to a practitioner (user) and allows the practitioner to determine the correctness of his / her pronunciation, and also allows the practitioner to acquire both the meaning and pronunciation of the word The purpose is to provide educational equipment that can do this.
[0007]
In addition, the present invention provides an educational device that allows a practitioner to acquire both the meaning and pronunciation of words by showing model characters and pictures and allowing the practitioner to answer in words. It is an object.
[0008]
[Means for Solving the Problems]
  In order to achieve the above object, the invention according to claim 1Is a record holding means for recording the model voice and the presentation information related to the model voice;
Presenting means for presenting information for presentation recorded in the record holding means;
Audio reproduction means for reproducing the exemplary audio recorded in the record holding means;
Voice input means for receiving the spoken voice;
Voice recognition for comparing whether the voice input by the voice input unit is similar to the model voice by comparing the voice input by the voice input unit with the model voice recorded in the voice recording unit Means,
When the voice recognition means recognizes that the voice input by the voice input means is similar to the model voice, the voice input by the voice input means or a feature pattern of the voice is used as the determination information as the determination information. Registration means for recording in the record holding means in association with the voice,
The voice input means further accepts a voice uttered based on the presentation information presented by the presentation means,
The speech recognition means further determines whether the uttered speech or the feature pattern of the speech is similar to the determination information stored in advance and determines whether the uttered speech is correct or incorrect.
  It is characterized by that.
[0009]
  The invention of claim 2In the educational device according to claim 1, when the voice recognition means recognizes that the voice inputted by the voice input means is not similar to the model voice, the voice reproduction means again outputs the model voice. PlayIt is characterized by that.
[0010]
  The invention of claim 3In the educational device according to claim 1 or 2, when the voice recognition means for judging the correctness of the uttered voice is judged to be an error, the voice corresponding to the presentation information is sent from the record holding means. Read and play back by the sound playback meansIt is characterized by that.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing a configuration example of an educational device according to the present invention. Referring to FIG. 1, this educational device includes a control unit 1 that controls the whole, a voice input unit (for example, a microphone) 2 that inputs voice, a voice recording and holding unit (for example, memory) 3 that records and holds voice, A sound reproducing unit (for example, a speaker) 4 that reproduces the sound recorded and held in the sound recording and holding unit 3 and an instruction unit (for example, a keyboard) 5 that indicates one of the sounds recorded in the sound recording and holding unit 3 The control unit 1 is provided with a voice recognition means 10 capable of registering a voice pattern for recognition (standard pattern) and capable of recognizing voice, and a comparison means 20, and is provided with a standard for recognition. At the time of pattern registration, the control unit 1 causes the audio reproducing unit 4 to reproduce the audio recorded and held in the audio recording holding unit 2, and when the user utters audio close to the reproduced audio, the user speaks. Based on the voice ( A standard pattern for recognition is generated and registered in the voice recognition means 10, and at the time of voice recognition, the control unit 1 determines the voice to be uttered as an instruction unit. 5, after the user is instructed by 5, the user utters the voice, the user's voice is recognized by the voice recognition means 10, and the recognition result is associated with the voice instructed by the instruction unit 5. The comparison means 20 determines whether or not and presents the result to the user.
[0018]
More specifically, the voice record holding unit 3 includes, for example, a control program for operating the educational device, pronunciation data of foreign words (for example, English voice data as an example), foreign language words. The meaning of Japanese is remembered.
[0019]
As the voice recognition method of the voice recognition means 10, any conventionally known method can be used. For example, the system described in the document “Furui Digital Audio Processing (Tokai Univ. Publishing, 1985)” can be used.
[0020]
FIG. 2 is a diagram showing a configuration example of the voice recognition means 10. In the example of FIG. 2, the speech recognition unit 10 includes a feature extraction unit 11, a specific speaker standard pattern registration unit 12, a comparison unit 13, and a result output unit 14.
[0021]
In the case where the speech recognition means 10 is configured as shown in FIG. 2, when a standard pattern is registered, an input speech of a predetermined word of a user (specific speaker) is inputted to a feature amount (feature pattern) by a feature extraction unit 11. ), The feature pattern extracted by the feature extraction unit 11 is stored in the standard pattern registration unit 12 for the specific speaker as a standard pattern. On the other hand, at the time of speech recognition, after the input speech of an unknown word of the user (specific speaker) is converted into a feature amount (feature pattern) by the feature extraction unit 11, the comparison unit 13 features the input speech of the unknown word. Similarities are calculated between the patterns and the standard patterns of various words registered in advance in the standard pattern registration unit 12 for specific speakers, and the result output unit 14 selects the standard pattern that gives the highest similarity. The words they have are output as recognition results.
[0022]
Next, an example of processing operation of the educational equipment having such a configuration will be described. First, a case where a standard pattern for specific speaker voice recognition is registered in the voice recognition means 10 will be described. Now, when the user designates (inputs) a word (word name) in Japanese, for example, from the instruction unit 5, the control unit 1 presents the Japanese to the user in the manner described above. As a way of presenting Japanese, for example, as shown in FIG. 1, a display unit 6 may be further provided, and Japanese characters (characters) may be displayed on the display unit 6. May be stored as waveform data in the audio recording / holding unit (memory) 3 and reproduced from the audio reproducing unit (speaker) 4. Note that the display unit 6 is not necessarily provided when Japanese is presented by voice reproduction.
[0023]
Thus, after presenting the Japanese language to the user, the control unit 1 reads out the pronunciation voice data of the foreign language word corresponding to the Japanese word from the voice record holding unit 3 and reproduces it. Thereby, the user can hear the pronunciation of the foreign language word corresponding to the Japanese word through the speaker 4. For example, it is assumed that a Japanese word and a corresponding foreign language (English) word are registered in the voice recording holding unit 3 with the number of words “5”. Here, the five words are “red”, “blue”, “green”, “white”, and “black”. First, when the user inputs “red” from the keyboard 5, the speaker 4 produces a pronunciation “red”. The user imitates this and pronounces “red” toward the microphone 2. The user may repeat this utterance a plurality of times as necessary.
[0024]
The sound (for example, “red”) input from the microphone 2 is A / D converted and provided to the control unit 1. In the control unit 1, a part of the given voice can be given to the voice recognition means 10, and the other part can be stored as a voice for reproduction in the voice record holding unit (memory) 3 as a voice waveform. In this example, the voice signal is branched and input to both the voice recognition means 10 and the memory 3, but it is not always necessary to branch the voice signal and input it to both. For example, a voice signal input from the microphone 2 may be stored in the memory 3 as a voice waveform and input to the voice recognition unit 10 as necessary. Further, the voice signal may be stored in the memory 3 as the voice waveform. However, when it is not necessary to reproduce the user's voice, the voice signal is given only to the voice recognition means 10 and converted into a feature pattern for voice recognition. It may be memorized. In this case, the amount of memory used can be reduced.
[0025]
As described above, when the voice uttered by the user is input to the voice recognition unit 10, the voice recognition unit 10 registers the input voice (for example, “red”) as a standard pattern for specific speaker voice recognition. Available to: That is, the feature pattern of the input voice (for example, “red”) can be registered in the standard pattern registration unit 12 for the specific speaker as a standard pattern for recognition of the user (specific speaker).
[0026]
Such an operation is repeated in sequence to practice pronunciation of all five words, create a standard pattern for this speaker based on the English utterances of the five words, and register this in the speech recognition means 10. . When the pronunciation practice for all five words is completed, the user can select a test mode from the instruction unit (keyboard) 5 and specify a word name. When the test mode is selected and the word name is designated, the control unit 1 reproduces the Japanese (designated Japanese) stored in the voice record holding unit (memory) 3 with, for example, voice and plays back the user. When this reproduction is completed, the voice recognition means 10 enters a state of waiting for recognition with respect to an unknown input voice.
[0027]
Next, a description will be given of recognition processing for an unknown input speech, that is, actual speech recognition. First, when the user designates (inputs) a word (word name) in, for example, Japanese from the instruction unit 5, the control unit 1 presents the Japanese to the user in the manner described above. As a way of presenting Japanese, for example, as shown in FIG. 1, a display unit 6 may be further provided, and Japanese characters (characters) may be displayed on the display unit 6. May be stored as waveform data in the audio recording / holding unit (memory) 3 and reproduced from the audio reproducing unit (speaker) 4. Note that the display unit 6 is not necessarily provided when Japanese is presented by voice reproduction.
[0028]
In this way, after presenting Japanese to the user, the user can try to speak English of the word presented in Japanese from the voice reproduction unit 4. When the user utters this word in English, the voice uttered in English is taken into the control unit 1, and the control unit 1 gives this voice to the voice recognition means 10 for voice recognition. The speech recognition means 10 obtains the feature pattern of the input English speech, compares this feature pattern with the standard patterns of five words (English) registered in advance, and the word having the most similar standard pattern The recognition result is output from the result output unit 14. At this stage, the control unit 1 compares the previously specified word name with the recognition result by the comparison means 20, and if the result of this comparison is the same, the control unit 1 determines that the answer is correct. Judge. Then, for example, the determination result is displayed on the display unit 6 to notify the user. At this time, the display order of the words may be the registration order, the reverse order of registration, or may be random.
[0029]
3 and 4 are flowcharts for explaining a specific example of the processing operation of the educational device of FIG. In the example of FIGS. 3 and 4, the user first inputs a file name for recording personal information (step S1). The file name may be input by the user from the keyboard 5, or the device itself automatically reads the word name stored in advance in the device memory, and the device itself is automatically generated. Is also possible. When the file name is input in this way, it is determined whether or not the file name is a new file (step S2). As a result, since a new file is used for the first time, it is necessary to make the user practice pronunciation and create a standard pattern for speech recognition. For this purpose, first, the word counter WCNT is initially set to “1” (step S3), and a Japanese word name is input (step S4). Similarly to the input of the file name, the Japanese word name can be input from the keyboard 5 by the user, or the device itself automatically reads the word name stored in the device memory. It is also possible for the device itself to occur automatically.
[0030]
Next, an English word name corresponding to the Japanese word name is input (step S5). For example, when “red” is input as a Japanese word name in step S4, “red” is input in step S5. At this time, the English input may be performed manually, or the device may input automatically.
[0031]
Thus, when Japanese and English corresponding to it are input, these are displayed, for example, on the display unit 6 (step S6). That is, for example, “red” and “red” are displayed. Then, this educational device outputs an exemplary pronunciation, and causes the user to utter (return) the voice accordingly (step S7). That is, the device outputs “red” as a model voice, and the user utters “red” as closely as possible to the model voice.
[0032]
This voice uttered by the user is taken into the voice recognition means 10, and the voice recognition means 10 registers the feature pattern of the voice of the word uttered by the user as a standard pattern of this word, for example, in a file (step S8). Next, the word counter WCNT is incremented by “1” (step S9), and it is determined whether or not the counter value WCNT exceeds the total number of words n (step S10). If it does not exceed n, the processing from step S4 to step S9 is repeated until n is exceeded, and when the processing is completed, the file is saved.
[0033]
Next, the user is allowed to select whether to practice pronunciation only or to practice further memory training in addition to pronunciation practice (step S11). When the user just listens to the model voice and practice the pronunciation, the process is completed. However, when the pronunciation practice and the memory training of the word are performed, the test routine is entered from here (that is, the process proceeds to step S15).
[0034]
If it is determined in step S2 that the input file name already exists and the user has already practiced speaking, the English vocabulary and the Japanese vocabulary associated with each other are loaded (steps S12 and S13). ) The standard pattern (template) for speech recognition created in step S8 is loaded (step S14). Here, the vocabulary is a set of words.
[0035]
Next, the test routine from step S15 is entered. In the test routine, first, the word pointer is set at the first word position (step S15). Then, the word counter WCNT is initialized to “1” (step S16). Next, the Japanese word at the word position is presented (for example, displayed) on the display unit 6, for example (step S17). Here, it is not always necessary to display Japanese words as characters, and recorded words may be output.
[0036]
Thus, when presenting a Japanese word to the user, the user can utter the corresponding English word (step S18). When the user utters an English word and the voice is input, the voice recognition means 10 recognizes the voice of the English word (step S19). That is, the voice feature pattern of the English word is extracted, and the feature pattern is collated with the standard pattern of each word registered in step S8, thereby performing speech recognition. Then, as a result of the speech recognition, it is determined whether or not the feature pattern of the speech uttered by the user has a correct speech feature amount (standard pattern) of the English word (step S20), and the utterance of the user is determined. If the received speech does not have the correct English word speech feature (standard pattern) (if it has been rejected or is a misrecognition result), the process returns to step S17 to display the Japanese word again and Then, the corresponding English word is uttered again, and the speech recognition processing in steps S17 to S20 is repeated.
[0037]
On the other hand, if the recognition result is correct in step S20, the word counter WCNT is incremented by “1” (step S21), and it is determined whether or not the word counter WCNT exceeds a predetermined value n (step S22). As a result, when the predetermined value n has not been reached, the process returns to step S17 again, the next Japanese word is displayed, and the test routine is repeated. In this way, when the word counter WCNT exceeds the predetermined value n in step S22, all processing is completed.
[0038]
In the above test routine (steps S15 to S22), the word pointer is set at the first word position, and the test is performed sequentially from the word at the first word position. Instead of this, for example, Alternatively, a random number may be generated to randomly determine a word to be tested.
[0039]
By performing such a test, the user can learn the correct pronunciation of the English word corresponding to the Japanese word name, and at the same time, the meaning of the English word (ie, the Japanese word) You can figure out what it is. In the above example, a word is presented to the user. However, a Japanese sentence may be presented, and an English sentence corresponding to the Japanese sentence may be presented to the user. In the above-described example, a keyboard is used as the instruction unit 5, but a recording medium such as a floppy disk or a CD-ROM can be used instead of the keyboard. In the above example, the instruction unit 5 is provided, but instead of the instruction unit 5, a voice recording unit (in which voices of contents uttered in two or more languages are recorded in association with each other ( A recording medium such as a floppy disk or a CD-ROM can also be provided.
[0040]
FIG. 5 is a diagram showing another configuration example of the educational device according to the present invention. The educational device of FIG. 5 is uttered in two or more languages in place of the instruction unit 5 in the educational device of FIG. An audio recording unit 7 (such as a floppy disk or a recording medium such as a CD-ROM) is provided in which audio of the contents described above is recorded in association with each other.
[0041]
In the educational device of FIG. 5, the voice recording holding unit 3 also has a function as a temporary storage unit that temporarily stores the content recorded in the voice recording unit 7, and at the time of registering a standard pattern for recognition, For a certain word, the control unit 1 transmits voices in one or more languages (for example, English voices) out of two or more language voices recorded in the voice recording holding unit (temporary storage unit) 3 from the voice recording unit 7. ) As a first voice from the voice playback unit 4, causing the user to utter English voice according to the first voice (English voice), and based on the voice uttered by the user, a standard pattern for recognition Is registered in the voice recognition means 10, and at the time of voice recognition, the control unit 1 from the voice recording unit 7 to the voice recording holding unit (temporary storage unit) 3 for a certain word. From two or more languages recorded in The voice of a language different from the kind of language spoken by the user (for example, Japanese voice) is reproduced as the second voice, and the second voice (Japanese voice) is transmitted to the user. The first voice corresponding to the voice (English voice) is uttered, the voice of the user is recognized by the voice recognition means 10, and the recognition result is recorded in the voice record holding unit (temporary storage unit) 3 and the first voice is recorded. It is determined whether or not it is associated with the voice reproduced as the voice, and presented to the user.
[0042]
In the educational device of FIG. 5, various contents can be set as the contents of the sound recording unit (recording medium) 7, and this content is temporarily stored in the sound recording holding unit 3, so By changing the type of word to be promoted, changing the type of language, and changing the program, it is possible to teach not only foreign languages but also answers to questions and to be a training machine for visually impaired people . Therefore, the function of the educational device can be easily changed by replacing only the voice storage unit 7, that is, the recording medium.
[0043]
In this way, in the educational device of FIG. 5, by replacing the recording medium, one person can use one system for various levels of learning.
[0044]
By the way, in each of the above-described educational devices (more specifically, a device in which the voice recognition means 10 of the control unit 1 has the configuration shown in FIG. 2), the user's error cannot be pointed out from the system side. Specifically, if there is an error that the person does not notice, for example, if there is a person who always pronounces “red” as “let”, and pronounces “let” along the guidance of “red”-“red”, If this is registered as a standard pattern of this specific speaker and the question “What does red say?” Is pronounced as “let” in the test, the result of speech recognition will be correct. For this reason, the above-mentioned educational equipment cannot correct the pronunciation error until someone points out that the pronunciation of the model is different from the pronunciation of the person.
[0045]
FIG. 6 is a diagram showing a configuration example of the speech recognition means 10 ′ intended to enable correction for an error that is pronounced with the user's belief. That is, in the configuration example of FIG. 6, it is intended to create a standard pattern with the correct pronunciation as much as possible and to learn the correct pronunciation. The speech recognition means 10 ′ includes the feature extraction unit 11, In addition to the specific speaker standard pattern registration unit 12, the comparison unit 13, and the result output unit 14, it further includes a non-specific speaker standard pattern registration unit 15.
[0046]
The configuration example in FIG. 6 uses two features: a voice recognition device for unspecified speakers is now available, and the recognition accuracy of the specific speaker method is higher. In the configuration example of FIG. 6, first, the standard pattern for unspecified speakers is used to check whether the user has made a correct pronunciation, and the specified speaker is determined to be correct. A standard pattern is registered.
[0047]
FIG. 7 is a diagram showing another configuration example of the educational equipment when the voice recognition means of the control unit 1 has the construction of the voice recognition means 10 ′ as shown in FIG. 6, and the educational equipment in FIG. In this way, it is intended to make it possible to correct mistakes that are pronounced with the user's assumptions. That is, the educational device of FIG. 7 is intended to create a standard pattern with the correct pronunciation as much as possible and to learn the correct pronunciation.
[0048]
The educational device in the example of FIG. 7 includes the instruction unit in addition to the voice recognition unit of the control unit 1 having the configuration of the voice recognition unit 10 ′ as shown in FIG. Along with (for example, a keyboard) 5, an audio recording unit (recording medium) 7 as shown in the configuration example of FIG. 5 is further provided. Here, in the voice recording unit (recording medium) 7, voices of contents uttered in two or more languages are recorded in association with each other. At this time, the contents of the uttered contents to be recorded are recorded. The voice is that of an unspecified speaker (for example, that of a standard voice obtained by averaging the voices of a plurality of speakers).
[0049]
In the educational device having the configuration shown in FIGS. 6 and 7, prior to registering the standard pattern of the user's voice in the standard pattern registration unit 12 for the specific speaker, the feature pattern of the voice uttered by the user and the unspecified speaker The similarity with the standard pattern for unspecified speakers registered in the standard pattern registration unit 15 is obtained, and whether or not a correct recognition result is obtained is checked. If the pattern is registered as it is in the standard pattern registration unit 12 for the specific speaker and a correct recognition result cannot be obtained, a message such as “Let's practice speaking again” or “Is the pronunciation correct?” Is given to the user. And repeat the same operation as described above. By performing such an operation, the feature pattern of the user voice that has the highest similarity with the standard pattern for the unspecified speaker registered in the standard pattern registration unit 15 for the unspecified speaker is used for the specific speaker. Can be registered in the standard pattern registration unit 12 for the specific speaker.
[0050]
Specifically, in order to check whether or not the user has made a correct pronunciation, when the voice of a certain word is uttered, the voice recognition means 10 'extracts the feature pattern of the inputted word and inputs it. First, the speech feature pattern is compared with the standard pattern for the first unspecified speaker. Then, the similarity between the two at that time and the word name of the standard pattern that gave this similarity are temporarily stored in, for example, a memory (not shown), and then the feature pattern of the input speech is used for the next unspecified speaker. Compare with the standard pattern. When the similarity to this standard pattern is greater than the similarity to the previous standard pattern, the previously stored standard pattern is deleted, and the current similarity and the word name of the standard pattern that gave the similarity are Store in memory. On the other hand, when the current similarity is smaller, the previously stored standard pattern is stored and held in the memory as it is. In this way, after the feature pattern of the input speech is sequentially compared with the standard patterns for unspecified speakers, and these similarities are obtained, the standard pattern giving the highest similarity, that is, the memory remains in the memory. The word with the highest similarity (word name) is the recognition result.
[0051]
When the recognition result is obtained from the voice recognition unit in this way, the control unit 1 compares the recognition result with the word name that urges the user to speak, and the word names match. If so, it is judged as correct pronunciation, and if it is different, it is judged as wrong pronunciation. This makes it possible to point out pronunciation errors that the person does not notice. When the recognition result matches the word name that prompted the user to speak, the standard pattern for the unspecified speaker that gave the recognition result can be registered as the standard pattern for the specific speaker. it can.
[0052]
As described above, in the educational device having the configuration of FIG. 6 and FIG. 7, prior to registering the standard pattern of the user's voice in the standard pattern registration unit 12 for the specific speaker, The similarity to the standard pattern for unspecified speakers registered in the standard pattern registration unit 15 for unspecified speakers is obtained, and whether or not a correct recognition result can be obtained is checked, the user pronounces himself correctly. Can use the educational equipment with the correct voice. At the same time, you can acquire the correct pronunciation yourself. That is, it is possible to correct an error that is pronounced with the user's belief, to create a standard pattern with the correct pronunciation as much as possible, and to learn the correct pronunciation.
[0053]
Also, in each of the above educational devices, if the speech recognition result is incorrect during learning such as word utterance, forgetting the pronunciation when presented by the device and uttering a completely different word. There are cases where the pronunciation is similar to the suggested pronunciation but is different from the correct pronunciation when registered. In any case, the user needs to hear the correct pronunciation again.
[0054]
FIG. 8 is a diagram illustrating another configuration example of the educational device according to the present invention, and the educational device of FIG. 8 is intended to solve the above-described problem.
[0055]
That is, referring to FIG. 8, in this educational device, the voice recognition means of the control unit 1 has a structure of voice recognition means 10 ″ as shown in FIG. 9, for example. In addition, in the example of the educational device in FIG. 8, an audio recording unit (recording medium) 7 is provided in addition to the instruction unit 5.
[0056]
8 and 9, when it is determined that the recognition result of the voice recognition unit 10 '' is different in the comparison unit 20, the control unit 1 makes one or both of a pair with the instructed voice. Are read from the voice recording / holding unit (temporary storage unit) 3 and given to the voice reproduction unit 4 to be reproduced.
[0057]
Specifically, also in the educational device of FIG. 8, the voice recording holding unit 3 receives, for example, a control program for operating the educational device from the voice recording unit (recording medium) 7, and the pronunciation sound of a foreign word. Data (for example, English voice data as an example), Japanese meanings of foreign words, etc. are loaded and stored.
[0058]
Also in this educational device, at the time of registration of the standard pattern, the voice of the word held in the voice recording holding unit 3 is played back from the voice playback unit 4, and the voice close to the played back voice is uttered to the user repeatedly. The feature pattern is registered in the standard pattern registration unit 12 for the specific speaker as a standard pattern for the specific speaker. After that, Japanese meaning English to be uttered is displayed, and the English pronunciation uttered is recognized in the same manner as described above to obtain a recognition result. At this time, it is also effective to temporarily store the spoken English voice in, for example, the voice recording holding unit 3.
[0059]
By the way, in this educational equipment, when misrecognizing as a result of such recognition, the control part 1 takes out the English sound of the applicable word from the voice record holding part 3, and this voice signal is used as the voice reproduction part 4 Play from and let the user hear. Subsequently, the control unit 1 reproduces the user's pronunciation voice temporarily stored in the voice record holding unit 3 from the voice reproduction unit 4 to let the user hear it. As a result, the user can clearly grasp the difference between the correct English pronunciation and the pronunciation uttered by the user. In other words, in this type of educational device, if the speech recognition result is incorrect during learning such as word utterance, forgetting the pronunciation when the pronunciation is presented from the device and uttering a completely different word. There are cases where the pronunciation is similar to the suggested pronunciation, but may be different from the correct pronunciation when registered, but with the educational device in FIG. Can hear the correct pronunciation again.
[0060]
FIG. 10 is a diagram showing a modification of the educational device of FIG. 8. The educational device of FIG. 10 uses the speech recognition means 10 of FIG. 2 as the speech recognition means of the control unit 1 in the educational device of FIG. Yes. That is, in the educational device of FIG. 10, when the recognition result of the voice recognition means 10 is different, the control unit 1 reproduces the instructed voice again to ask the user to speak, and the user When the voice is uttered again, the standard pattern for voice recognition previously registered is rewritten with the feature pattern of the voice.
[0061]
That is, when using educational equipment, the user may have to speak a language that he / she does not know. For this reason, the utterance is not stable or is mistaken. Of these, in order to reduce the instability of utterance, it is effective to use this educational equipment repeatedly, which can stabilize the utterance, but the error is to rewrite the original standard pattern There is a need.
[0062]
In addition to the above causes, there is a change over time in voice recognition errors. In other words, when time passes after the voice is registered, there is a case where correct recognition cannot be performed even though the correct pronunciation is made.
[0063]
In the educational device in FIG. 10, the misrecognized voice standard pattern is replaced with a new one, so the above case can be dealt with.
[0064]
Specifically, also in the educational device of FIG. 10, the voice recording holding unit 3 receives, for example, a control program for operating the educational device from the voice recording unit (recording medium) 7, and the pronunciation sound of foreign words. Data (for example, English voice data as an example), Japanese meanings of foreign words, etc. are loaded and stored.
[0065]
Also in this educational device, at the time of registration of the standard pattern, the voice of the word held in the voice recording holding unit 3 is played back from the voice playback unit 4, and the voice close to the played back voice is uttered to the user repeatedly. Then, it is registered in the standard pattern registration unit 12 for specific speakers as a standard pattern for specific speakers. After that, Japanese meaning English to be uttered is displayed, and the English pronunciation uttered is recognized in the same manner as described above to obtain a recognition result. At this time, it is also effective to temporarily store the spoken English voice in, for example, the voice recording holding unit 3.
[0066]
By the way, in this educational equipment, when misrecognizing as a result of such recognition, the control part 1 takes out the English sound of the applicable word from the voice record holding part 3, and this voice signal is used as the voice reproduction part 4 Play from and let the user hear.
[0067]
At the same time, the voice recognition means 10 is set to the registration mode. Therefore, when the user utters a voice, the voice is voiced, and the feature pattern is registered as a standard pattern in the standard pattern registration unit 12 for the specific speaker. By registering the feature pattern as a standard pattern in this way, the existing standard pattern that has been previously registered, that is, the standard pattern that is currently erroneously recognized, is deleted. However, the existing standard pattern does not necessarily have to be erased and rewritten, and a new standard pattern obtained by averaging the existing standard pattern may be registered as a standard pattern. This can prevent the standard pattern from aging.
[0068]
Further, in the above-described educational device of each configuration example, when the voice recognition means misrecognizes, it cannot be distinguished whether the user uttered the wrong word or the correct word.
[0069]
FIG. 11 is a diagram showing another example of the configuration of the educational device according to the present invention. If the educational device of FIG. 11 misrecognizes speech recognition means, this is what the user uttered the wrong word. It is intended to distinguish whether the correct word is spoken.
[0070]
That is, the educational device of FIG. 11 has a configuration in which the control unit 1 is configured as shown in FIG. 12, for example, in the configuration examples of FIGS. Referring to FIG. 12, the control unit 1 of the educational device in FIG. 11 includes a speech recognition means 10 ′ ″, a feature extraction unit 11, a specific speaker standard pattern registration unit 12, a comparison unit 13, and a result output. In addition to the unit 14, the designated word similarity that holds the similarity obtained by the comparison unit 13 (similarity with respect to the standard pattern of the feature pattern of the speech of the word given the recognition result when the recognition result is obtained) A degree holding part 17 is provided. Further, the comparison unit 20 ′ of the control unit 1 compares the similarity held in the designated word similarity holding unit 17 of the speech recognition unit 10 ′ ″ with the threshold TH, and the similarity is larger or smaller than the threshold TH. It has come to judge.
[0071]
In the educational device having such a configuration, at the time of speech recognition, the control unit 1 instructs the user to utter the speech uttered by the instruction unit 5, for example, and then causes the user to utter the speech, The voice recognition means 10 ′ ″ recognizes the result, determines whether the recognition result is associated with the voice instructed by the instruction section 5, and then designates the voice to be uttered by the voice recognition means 10 ′ ″. The similarity is calculated between the feature pattern of the voice uttered by the user and the standard pattern, and the recognition result is obtained in the same manner as described above. As a result, when a correct recognition result is obtained, the same operation as described above is performed. On the other hand, when the recognition is not correctly performed, the calculated similarity is smaller or larger than a predetermined threshold TH. If it is smaller than the threshold value TH, the misrecognized voice is reproduced.
[0072]
Specifically, as a case where the voice recognition means misrecognizes, as described above, the user forgets the pronunciation when presented from the educational device and speaks a completely different word, or Although the pronunciation is similar, it is possible that the pronunciation is different from the correct pronunciation at the time of registration. In the former case, since the degree of similarity is lower than that of the latter, the two can be distinguished by the difference in degree of similarity. That is, when the recognition is not correctly performed, it is possible to distinguish the two by determining whether the calculated similarity is smaller or larger than a predetermined threshold TH. When the distinction is made in this way, in the former case, for example, a message “Is the word wrong?” Is shown to the user, and in the latter case, the message is “Different from this word. The message “Sho” is shown to the user, the voice of the misrecognized word is played, and the correct voice is played by saying “The correct answer is this”.
[0073]
The operation of the educational device in FIG. 11 will be described more specifically. In the educational device of FIG. 11 as well, the voice recording holding unit 3 is supplied from the voice recording unit (recording medium) 7, for example, a control program for operating the educational device, voice pronunciation data of a foreign language word (for example, a model) English voice data) and the meaning of Japanese words in foreign languages are loaded and stored.
[0074]
Also in this educational device, at the time of registration of the standard pattern, the voice of the word held in the voice recording holding unit 3 is played back from the voice playback unit 4, and the voice close to the played back voice is uttered to the user repeatedly. Then, it is registered in the standard pattern registration unit 12 for specific speakers as a standard pattern for specific speakers. After that, Japanese meaning English to be uttered is displayed, and the English pronunciation uttered is recognized in the same manner as described above to obtain a recognition result. At this time, it is also effective to temporarily store the spoken English voice in, for example, the voice recording holding unit 3.
[0075]
By the way, in this educational device, when the user utters the sound of a certain word in order to check whether he / she is pronounced correctly, the speech recognition means 10 ′ ″ has the characteristics of the sound of the input word. The pattern is extracted, and the feature pattern of the input speech is first compared with the standard pattern for the first unspecified speaker. Then, the similarity between the two and the word name of the standard pattern to which the similarity is given are temporarily stored in the designated word similarity holding unit 17, and the feature pattern of the input speech is then stored as the standard for the next specific speaker. Compare with pattern. When the similarity to this standard pattern is greater than the similarity to the previous standard pattern, the previously stored standard pattern is deleted, and the current similarity and the word name of the standard pattern to which the similarity is given are stored. To do. On the other hand, when the current similarity is smaller, the current object is deleted as it is, and the next standard pattern is taken out. However, when the word name of the standard pattern to be collated is the same as that uttered from the storage unit through the voice reproduction unit 4, it is stored in the same storage unit regardless of the similarity.
[0076]
In this way, after comparing the feature pattern of the input speech with each standard pattern for a specific speaker in order, and obtaining these similarities, the standard pattern giving the highest similarity, that is, remaining in the storage unit. The word with the highest similarity (word name) is the recognition result. If this recognition result is incorrect, the similarity stored together with the correct word name is compared with the threshold value TH. If the similarity is lower than the threshold TH, it is displayed with a message saying “Is the word wrong?”, Whereas if it is higher than the threshold TH, it is displayed with a message “Is it wrong?” The voice of the misrecognized word is taken out from the voice record holding unit 3 and output from the voice reproduction unit 4.
[0077]
As a result, the user (speaker) notices that he / she is wrong. Or, by knowing a word whose pronunciation is easy to be mistaken, the pronunciation is not mistaken.
[0078]
As a method for determining the threshold value TH, a value about 1/2 to 2/3 of the similarity generated between the standard pattern of the specific speaker method and the correct input voice feature pattern is appropriate.
[0079]
In this way, when the speech recognition result is misrecognition, it is possible to distinguish whether the user is uttering the wrong word or the correct word and to inform the user Become.
[0080]
Moreover, FIG. 13 is a figure which shows the other structural example of the educational equipment which concerns on this invention, and the educational equipment of FIG. 13 shows an example character and a picture, and makes a practitioner answer in words, The purpose is to provide educational equipment that allows students to learn both the meaning and pronunciation of words.
[0081]
That is, in the educational device of FIG. 13, for example, in the educational device of FIGS. 1 and 5, an audio / image recording holding unit (memory) 23 is provided instead of the audio recording holding unit 3. The holding unit 23 stores a program for operating the apparatus and foreign language pronunciation voice data of a word, and images (characters and pictures) corresponding to the program. Note that the example of FIG. 13 corresponds to FIG. 1 (instruction unit 5 is provided).
[0082]
In the configuration example of FIG. 13, for example, when a picture is specified from the instruction section (for example, keyboard) 5, the picture is displayed on the display section (display) 6, and voice data corresponding to this picture (foreign pronunciation) is voiced. Reading from the image record holding unit 23 and outputting the pronunciation of the foreign language from the sound reproduction unit (speaker) 4. Thus, the user can listen to the pronunciation of the foreign language corresponding to the picture through the speaker 4 while viewing the picture on the display unit 6. As an example, it is assumed that the foreign language is English and the number of words is five. Suppose now that the five words are “dog”, “cat”, “bird”, “horse”, and “cow”. When the program is started, a picture of the first dog is displayed on the display unit 6, and a pronunciation “dog” is output from the audio reproduction unit (speaker) 4. The user imitates this and pronounces “dog” toward the voice input unit (microphone) 2. If necessary, this may be repeated a plurality of times.
[0083]
The voice input from the voice input unit (microphone) 2 is A / D converted, partly input to the voice recognition means 10, and the other part is a voice / image record holding unit (remaining a voice waveform for reproduction) It may be stored in the memory 23). In this part, it is not always necessary to branch the audio signal into both. The voice waveform may be stored in the voice / image record holding unit (memory) 23 as it is and input to the voice recognition means 10 if necessary. The amount of memory used can be reduced if the feature amount is corrected and stored. The voice recognition means 10 creates a standard pattern for speaker recognition by using the input voice for voice registration for specific speaker voice recognition.
[0084]
Repeating these actions, all five words are pronounced. When a test mode is selected from the instruction unit (keyboard) 5, an animal picture is displayed on the display unit (display) 6, and a state of waiting for recognition is entered.
[0085]
The user tries to speak English on the picture displayed on the display unit 6. The voice uttered in English is input to the voice recognition means 10 and recognized among the five registered words, and the recognition result is output to the comparison means 20. Here, the word name sent earlier is compared with the recognition result. The display unit 6 informs the user of the result. In this way, pictures can be sequentially displayed on the display unit 6, and the user can sequentially speak the corresponding foreign language, and the result of whether or not the pronunciation is correct can be reported to the user. At this time, the display order of the pictures may be the registration order, the reverse order of the registration, or random.
[0086]
In this way, with the educational equipment of FIG. 13, it is possible to show the model characters and pictures and have the practitioner answer in words, so that the practitioner can learn both the meaning and pronunciation of the words. Become.
[0087]
In the configuration example of FIG. 13, the voice recognition means 10 has the same configuration as that shown in FIG. 2, for example. The input voice is converted into a feature amount by the feature extraction unit 11, and the voice is registered. The converted feature amount is directly stored in the standard pattern recording unit 12 for the specific speaker. On the other hand, at the time of recognition, after the unknown input speech is converted into a feature amount by the feature extraction unit 11, the comparison unit 13 calculates a similarity with each of the standard patterns registered in advance, and the most similarity With a higher value is output from the result output unit 14 as a recognition result.
[0088]
In the above example, the foreign language is English and the number of words is 5. However, the foreign language may be other than English, and the number of words can be any number. In the above example, the correspondence between the picture and the word in the foreign language is described. However, the present invention is not limited to this. For example, the national flag and the country name, the company emblem and the company name, the face and the person name, the kanji and the reading, the position on the map It is also possible to associate a location name with a location name. In addition, if a picture is converted into a moving picture, it is effective for education of correspondence between visual information and auditory information, such as learning sign language.
[0089]
In the above example, the word is used, but it goes without saying that it may be a sentence. In FIG. 13, in the above example, the instruction unit 5 is a keyboard, and commands are input and selected from the keyboard. However, the instruction unit 5 is not necessarily a keyboard, for example, a floppy disk or the like. The program may be used and controlled by a program stored on the floppy disk.
[0090]
【The invention's effect】
  As explained above, claims 1 toClaim 3According to the described invention, it is possible to provide an educational device that shows an exemplary voice and judges the correctness of the pronunciation of the practitioner, and can acquire both the meaning of the word and the pronunciation at once.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration example of an educational device according to the present invention.
FIG. 2 is a diagram showing a configuration example of voice recognition means.
FIG. 3 is a flowchart for explaining a processing operation of the educational device in FIG. 1;
4 is a flowchart for explaining a processing operation of the educational device in FIG. 1. FIG.
FIG. 5 is a diagram showing another configuration example of the educational device according to the present invention.
FIG. 6 is a diagram showing another configuration example of voice recognition means.
FIG. 7 is a diagram showing another configuration example of the educational device according to the present invention.
FIG. 8 is a diagram showing another configuration example of the educational device according to the present invention.
FIG. 9 is a diagram showing another configuration example of voice recognition means.
10 is a diagram showing a modification of the educational device in FIG.
FIG. 11 is a diagram showing another configuration example of the educational device according to the present invention.
12 is a diagram illustrating a configuration example of a control unit of the educational device in FIG. 11. FIG.
FIG. 13 is a diagram showing another configuration example of the educational device according to the present invention.
[Explanation of symbols]
1 Control unit
2 Voice input part
3 Voice record holding part
4 Audio playback part
5 indicator
6 Display section
7 Voice recording part
10, 10 ′, 10 ″, 10 ′ ″ speech recognition means
11 Feature extraction unit
12 Standard pattern registration section for specific speakers
13 Comparison part
14 Result output section
15 Standard pattern registration section for unspecified speakers
20,20 'comparison means
23 Audio / Image Record Holding Unit

Claims (3)

Record holding means for recording the model voice and the presentation information related to the model voice;
  Presenting means for presenting information for presentation recorded in the record holding means;
  Audio reproduction means for reproducing the exemplary audio recorded in the record holding means;
  Voice input means for receiving the spoken voice;
  Voice recognition for comparing whether the voice input by the voice input unit is similar to the model voice by comparing the voice input by the voice input unit with the model voice recorded in the voice recording unit Means,
  When the voice recognition means recognizes that the voice input by the voice input means is similar to the model voice, the voice input by the voice input means or a feature pattern of the voice is used as the determination information as the determination information. Registration means for recording in the record holding means in association with the voice,
  The voice input means further accepts a voice uttered based on the presentation information presented by the presentation means,
  The speech recognition means further determines whether the uttered speech or the feature pattern of the speech is similar to the determination information stored in advance and determines whether the uttered speech is correct or incorrect.
  Educational equipment characterized by that. 2. The educational device according to claim 1, wherein when the voice recognition unit recognizes that the voice input by the voice input unit is not similar to the model voice, the voice playback unit reproduces the model voice again. Educational equipment characterized by that. In the educational device according to claim 1 or 2, when the voice recognition means for judging correctness of the uttered voice is judged to be an error, the voice corresponding to the presentation information is read from the record holding means, Educational equipment, which is played back by the voice playback means.

Images