JP2010204442A - Speech recognition device, speech recognition method, speech recognition program and program recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech recognition device capable of giving trigger that a user may notice that an error of a speech recognition result is possibly caused by a recognition mode error. <P>SOLUTION: When at least one or more recognition modes for performing speech recognition on an input speech, are selected in a plurality of recognition modes equipped beforehand, and a user indicates it as a first recognition mode from an input means 20, a recognition mode setting means 11 sets it to a condition defined by the first recognition mode, and a speech recognition means 12 performs recognition processing of the input speech based on the condition, and outputs the speech recognition result to the user via an output means 14. Then, a mode is changed to a second recognition mode which is different from the first recognition mode indicated by the user, and feed-back information for asking the user whether or not, speech recognition is performed again on the same speech data with the input speech data, is created at predetermined timing by a feed-back creation means 13, and output the information to the user via the output means 14. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

The present invention relates to a voice recognition device, a voice recognition method, a voice recognition program, and a program recording medium, and in particular, when a user has a function of designating a recognition mode for performing voice recognition, the user erroneously uses the designation of the recognition mode. The present invention relates to a voice recognition apparatus, a voice recognition method, a voice recognition program, and a program recording medium having a mechanism for performing feedback when the program is performed.

  Conventionally, there has been a speech recognition system that recognizes speech input by a user after setting the recognition mode designated by the user. For example, when there are recognition modes of “male” and “female” and the user designates “male”, the speech recognition system performs speech recognition processing on the input speech using a male-dependent sex-dependent acoustic model. Also, for example, there are recognition modes of “Japanese” and “English”, and when the user designates “Japanese”, the speech recognition system performs speech recognition processing on the input speech using a Japanese language model / word dictionary. To do.

  On the other hand, for example, as disclosed in Japanese Patent Application Laid-Open No. 07-104780 “Unspecified Speaker Continuous Speech Recognition Method” of Patent Document 1, a plurality of gender-dependent acoustic models for men and gender-dependent acoustic models for women are used. There is also a speech recognition system that executes recognition processing using an acoustic model at the same time. As described above, a speech recognition system that performs recognition processing using a plurality of models at the same time eliminates the need for the user to select and set a recognition mode, but does not result in recognition using a model that should be originally applied. There are cases where the recognition accuracy is degraded. In particular, when used in an environment in which resources such as a terminal and resources such as a CPU are highly restricted, a simultaneous recognition technique that requires a large amount of memory and processing power cannot be used. A speech recognition system that specifies and uses a recognition mode is preferable.

JP 07-104780 A (page 3-5)

  However, in the voice recognition system that uses the recognition mode as described above, the user may use the voice recognition system by mistakenly specifying the recognition mode. If the designation is used incorrectly, a completely different recognition result is often output.

  Furthermore, even if the specification is used incorrectly, for example, depending on the type of recognition mode, such as gender, the error in the specified recognition mode is the cause of erroneous recognition as the cause of the output of an incorrect speech recognition result. It is difficult for the user to notice. As a result, it is often the case that the user thinks that there was a problem with his / her voice, and that he / she utters again to perform voice recognition again. As long as the recognition mode is specified incorrectly, if the recognition mode is not set correctly, an incorrect recognition result will be output again, leading to a result of repeated useless operations.

(Object of the present invention)
The present invention has been made in view of such circumstances, and an object of the present invention is to return the appropriate feedback to the user, thereby recognizing the error in the recognition mode designated by the cause of the error in the speech recognition result. An object of the present invention is to provide a voice recognition device, a voice recognition method, a voice recognition program, and a program recording medium that can give the user an opportunity to notice that there is a possibility.

  In order to solve the above-described problems, the speech recognition apparatus, speech recognition method, speech recognition program, and program recording medium according to the present invention employ the following characteristic configuration. The numbers (1) and (12) below correspond to the item numbers in the claims.

(1) The user selects at least one or more recognition modes for performing speech recognition on the input speech from a plurality of recognition modes provided in advance and designates the first recognition mode as the first recognition mode. The first recognition specified by the user in a speech recognition apparatus comprising speech recognition means for performing speech recognition on the input speech and outputting a speech recognition result to the user based on a condition defined by a recognition mode. Change to a second recognition mode different from the mode, and generate feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice, and generate the user at a predetermined timing. A speech recognition apparatus comprising feedback generation means for outputting to the above.
(12) The user selects at least one or more recognition modes for performing speech recognition on the input speech from a plurality of recognition modes provided in advance and designates the first recognition mode as the first recognition mode. A speech recognition method for performing speech recognition on the input speech based on conditions defined by a recognition mode and outputting a speech recognition result to the user, which is different from the first recognition mode designated by the user. Change to the second recognition mode, generate feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice, and output the feedback information to the user Voice recognition method.

  According to the speech recognition apparatus, speech recognition method, speech recognition program, and program recording medium of the present invention, the following effects can be obtained.

  Not only the voice recognition result, but also feedback information that asks whether or not to perform voice recognition again after resetting the recognition mode is returned to the user, so the cause of the error in the voice recognition result may be an error in the specified recognition mode It is possible to make the user aware that there is

It is a block block diagram which shows an example of the internal structure of the speech recognition apparatus which concerns on this invention. It is a flowchart which shows an example of operation | movement of the speech recognition apparatus shown in FIG. It is a block block diagram which shows an example of the whole structure of the speech recognition apparatus which concerns on this invention. It is a block block diagram which shows the other example of the whole structure of the speech recognition apparatus which concerns on this invention. It is a block block diagram which shows the further different example of the whole structure of the speech recognition apparatus which concerns on this invention. It is a block block diagram which shows the further different example of the whole structure of the speech recognition apparatus which concerns on this invention.

  Preferred embodiments of a speech recognition apparatus, speech recognition method, speech recognition program, and program recording medium according to the present invention will be described below with reference to the accompanying drawings. In the following description, the speech recognition apparatus and speech recognition method according to the present invention will be described. However, the speech recognition method may be implemented as a speech recognition program that can be executed by a computer, or speech recognition. Needless to say, the program may be recorded on a computer-readable recording medium.

(Features of the present invention)
Prior to the description of the embodiments of the present invention, an outline of the features of the present invention will be described first. The speech recognition apparatus according to the present invention uses speech recognition means for performing speech recognition processing on input speech using a first recognition mode designated by a user, and again using a second recognition mode different from the first recognition mode. Feedback generating means for generating feedback information for inquiring whether or not to perform voice recognition processing, and providing feedback that triggers the user to check not only the voice recognition result but also the recognition mode. It is characterized in that information can be returned.

(Configuration example of embodiment)
Next, an example of the configuration of the speech recognition apparatus according to the present invention will be described with reference to FIG. FIG. 1 is a block diagram showing an example of the internal configuration of a speech recognition apparatus according to the present invention, and shows only the blocks related to the present invention.

  Referring to FIG. 1, a speech recognition apparatus 10 showing an example of the present invention includes at least a recognition mode setting unit 11, a speech recognition unit 12, a feedback generation unit 13, and an output unit 14. The recognition mode setting means 11 sets so as to use the recognition mode (first recognition mode) selected by an external user operation. The speech recognition unit 12 performs speech recognition processing on the input speech by using a speech recognition model or the like corresponding to the recognition mode (first recognition mode) set by the recognition mode setting unit 11. The word string as the recognition result is supplied to the output means 14.

  The feedback generation unit 13 uses a speech recognition model or the like corresponding to a recognition mode (second recognition mode) different from the recognition mode (first recognition mode) set by the recognition mode setting unit 11. Then, feedback information for inquiring whether or not to perform speech recognition processing again on the same input speech is generated and supplied to the output means 14. The output unit 14 outputs a word string as a recognition result supplied from the voice recognition unit 12 to the user. Further, the feedback information supplied from the feedback generation means 13 is also output to the user.

  Next, the operation of the speech recognition apparatus 10 shown in FIG. 1 will be described in detail with reference to the flowchart of FIG. FIG. 2 is a flowchart showing an example of the operation of the speech recognition apparatus 10 shown in FIG.

  In the flowchart of FIG. 2, first, the recognition mode setting means 11 sets the recognition mode according to the content of the user operation (step S101). Thereafter, the user inputs voice (step S102).

  When speech input is detected, the speech recognition unit 12 performs speech recognition processing on the input speech using a model or the like corresponding to the recognition mode set by the recognition mode setting unit 11, and a word is obtained as a speech recognition result. A column is generated (step S103). When the word string is generated as the voice recognition result, the output unit 14 outputs the word string of the voice recognition result generated by the voice recognition unit 12 to the user (step S104).

  On the other hand, the feedback generation unit 13 uses a model or the like corresponding to a recognition mode (second recognition mode) different from the set recognition mode (first recognition mode), and again performs speech for the same input voice. Feedback information for inquiring whether or not to perform recognition processing is generated and output to the user (step S105).

  Next, it is checked whether or not there is another voice recognition processing instruction from the user who has received the word string of the voice recognition result by the voice recognition unit 12 and the feedback information by the feedback generation unit 13 (step S106). If there is a reaction from the user to change the recognition mode and perform the recognition process again (Yes in step S106), the recognition mode setting means 11 changes to a new recognition mode (second recognition mode). (Step S107). Thereafter, the user inputs voice again (step S108).

  When the voice input is detected again, the voice recognition unit 12 uses the model corresponding to the recognition mode newly set by the recognition mode setting unit 11 (second recognition mode), etc. Then, voice recognition processing is performed, and a word string is generated as a voice recognition result (step S109). When the word string is generated as the voice recognition result, the output unit 14 outputs the word string of the voice recognition result generated by the voice recognition unit 12 to the user again, and the feedback generated by the feedback generation unit 13 again. The operation of outputting information to the user is repeated.

  As described above, the speech recognition apparatus according to the present invention determines whether or not to perform recognition processing using a recognition mode (second recognition mode) different from the recognition mode (first recognition mode) specified by the user. Feedback information to be inquired is output to the user, and the recognition mode (first recognition mode) is obtained by giving feedback to the user whether or not to perform the recognition processing again after changing the recognition mode (first recognition mode). It is possible to make the user aware that an erroneous specification of (mode) may be the cause of the recognition result error.

  Next, a more detailed configuration of the speech recognition apparatus in FIG. 1 will be described with reference to FIG. FIG. 3 is a block configuration diagram showing an example of the overall configuration of the speech recognition apparatus according to the present invention. The circuit block (same as in FIG. 1) in the speech recognition apparatus shown in FIG. A recognition mode and input means 20 for inputting voice to the voice recognition device 10 are also shown. Note that the input unit 20 may be disposed in the voice recognition device 10 depending on circumstances.

  In the overall configuration of the speech recognition apparatus shown in FIG. 3, the input unit 20 supplies information on the recognition mode (first recognition mode) designated by the user to the recognition mode setting unit 11, and the input speech uttered by the user. Is provided to the voice recognition means 12 and the like, and includes, for example, a display screen, buttons, a keyboard, a microphone, and the like.

  The recognition mode setting means 11 in the speech recognition apparatus 10 shown in FIG. 3 is a recognition mode (first recognition mode) corresponding to a user operation input from the outside via the input means 20 as in FIG. ) To perform voice recognition processing. That is, the acoustic model, language model, word dictionary, etc. used in the speech recognition process of the speech recognition means 12 are set to those corresponding to the recognition mode (first recognition mode) designated by the user.

  Here, the input means 20 is provided with a mode designation button corresponding to the recognition mode, for example, and the user designates the recognition mode (first recognition mode) by pressing the mode designation button. When setting the recognition mode (first recognition mode), not only one recognition mode but also a plurality of recognition modes such as a recognition mode related to gender and a recognition mode related to language may be set at a time.

  As the recognition mode, for example, there are recognition modes relating to the genders of “male” and “female”, and when the user designates the recognition mode of “male”, the recognition mode setting unit 11 is more accurate for the voice input by the male. It is set to use a male-dependent sex-dependent acoustic model created so that it can be recognized. Also, there are recognition modes related to the languages “Japanese” and “English”, and when the user designates the recognition mode of “Japanese”, the recognition mode setting means 11 recognizes Japanese input speech with higher accuracy. It is set to use a language model or word dictionary created so that it can. Thus, models used in speech recognition processing such as an acoustic model, a language model, and a word dictionary correspond to the recognition mode.

The speech recognition unit 12 performs speech recognition processing on the input speech using the models set by the recognition mode setting unit 11, and generates a recognition result as a word string. For example, the voice recognition processing is performed by converting input voice to MFCC (Mel-Frequency
Cepstral Coefficient (mel frequency cepstrum coefficient) and other acoustic analysis processing that converts into speech features, speech detection processing that separates speech and silent segments from the input speech, and acoustic models and language models for speech features A search process for generating a word string similar to the input speech using models such as a word dictionary is used. However, since this is a technique well known to those skilled in the art, detailed description thereof is omitted.

  As described above, the feedback generation unit 13 changes to a recognition mode (second recognition mode) different from the recognition mode (first recognition mode) set by the recognition mode setting unit 11 and performs voice recognition processing again. Feedback information for inquiring whether or not is generated is supplied to the output means 14. For example, feedback information that asks the user whether or not to perform re-recognition by changing the recognition mode may be generated, such as "Do you want to change the recognition mode and try again?"

  As described above, the output unit 14 outputs a word string as a speech recognition result supplied from the speech recognition unit 12 to the user. Further, the feedback information supplied from the feedback generation unit 13 is output to the user.

  After the feedback information is output to the user, when the user reacts to process the voice recognition again, the recognition mode setting unit 11 resets to a new recognition mode (second recognition mode). Again, the speech recognition means 12 performs speech recognition processing on the input speech. Here, the reaction that the user processes voice recognition again is, for example, replying “Yes” to the output feedback information by voice, or the first designated recognition mode (first recognition mode) and Is performed by pressing a mode designation button of a different recognition mode (second recognition mode).

  In the case of reacting to process voice recognition again, for example, the user may change to the designated recognition mode by redesignating the recognition mode. Alternatively, in the alternative 2 recognition mode, a recognition mode (second recognition mode) different from the initially set recognition mode (first recognition mode) may be automatically selected and set. Alternatively, when a function for learning the error rate of the recognition result is provided, a recognition mode corresponding to the learning result may be selected and set as the second recognition mode.

  As described above, an error is included in the word string of the speech recognition result output immediately before by outputting feedback information for inquiring whether or not to perform the speech recognition processing again after changing the recognition mode. If the error has occurred, it is possible to give the user a notice that the cause of the error may be an error in the designated recognition mode.

(Other embodiments)
Next, an embodiment different from the above-described embodiment will be further described.

  The voice recognition device 10 includes voice storage means for storing voice data, and the voice recognition means 12 stores all or part of the input voice to be voice-recognized as voice data in the voice storage means. You may make it leave. If there is a reaction from the user to process the voice recognition again, the voice recognition process can be performed using the voice data stored in the voice storage means. Thereby, after changing the recognition mode, it is possible to reduce time and effort for the user to input voice again.

  When the voice recognition process can be performed using the voice data stored in the voice storage unit, the feedback generation unit 13 uses the recognition mode (first recognition mode) initially set by the recognition mode setting unit 11. The mode is changed to a recognition mode (second recognition mode) different from (mode), and feedback information for inquiring whether or not to perform voice recognition processing again on the voice data stored in the voice storage means is generated. For example, feedback information such as “Do you want to change the recognition mode and repeat the speech recognition process earlier?” May be generated.

  Regarding the storage of the voice data in the voice storage means, the input voice may be stored as it is in, for example, a waveform format, or may be stored in a format to which a conversion process such as a voice feature after acoustic analysis processing is added. It doesn't matter. In the case of storing in the voice storage means in the form of the voice feature amount, the acoustic analysis process can be omitted when the voice recognition process is performed again on the voice data. The amount of processing can be reduced.

  Further, all input voices may be stored in the voice storage means, or only the section determined as the voice section by the voice detection process may be stored. When only the section determined as the voice section is stored in the voice storage unit, the size of the voice data to be stored can be reduced, and the voice detection process is performed when the voice recognition process is performed again on the voice data. Therefore, it is possible to reduce the processing amount in the second speech recognition process.

  Also, feedback information for inquiring whether or not to perform the speech recognition process again is output to the user, and the recognition mode is changed prior to the timing at which a response to restart the speech recognition process is obtained from the user. Voice recognition processing may be performed on the voice data stored in the memory. In this case, it is possible to shorten the time from when a response indicating that the voice recognition process is performed again from the user until the result of the voice recognition process is obtained.

  Further, the voice recognition process is performed again on the feedback information generated by the feedback generation unit 13, that is, the voice previously input in the recognition mode (second recognition mode) different from the previous recognition mode (first recognition mode). The timing for outputting feedback information for inquiring whether or not to the user is not almost the same as the timing for outputting the word string of the speech recognition result to the user, and that the user has performed any of the following operations with the input means 20. It may be performed at the detected timing. The user performs the following operation with the input means 20 because the user sees the word string of the output voice recognition result, notices the setting error of the recognition mode, and sets the recognition mode again to try the voice recognition process again. May be specified. In some cases, the recognition mode is changed to try another utterance.

  Therefore, if the feedback information is output to the user at the timing when it is detected that the user has performed any of the following operations with the input means 20, the recognition mode is re-designated by recognizing the recognition mode setting error. Whether you are trying to change the recognition mode or trying to make another utterance, do the process of changing the recognition mode without the hassle of generating and outputting feedback information In addition, even if the user himself / herself notices an incorrect recognition mode specification and redoes voice recognition processing, it is possible to reduce the trouble of redoing voice input from the beginning after obtaining feedback information. It becomes.

  In addition, as feedback information generated by the feedback generation unit 13, feedback information including a recognition mode candidate to be changed is generated together with information for inquiring whether or not to perform speech recognition processing again on previously input speech. Anyway. For example, feedback information such as “Do you want to re-recognize by changing to the female recognition mode?” Is output to the user. In addition, the recognition mode candidate to be selected may be clearly presented to the user by blinking the mode designation button for the recognition mode candidate to be changed.

  For example, if the recognition mode candidate to be changed is a two-choice recognition mode, it may be determined that the recognition mode is not designated first by the user, or if the recognition mode is incorrect. An error pattern may be learned in advance and determined from the learned error pattern. As described above, when a user who has confirmed a recognition mode candidate newly designates a recognition mode, the recognition mode designated by the user may be used. This makes it possible to clarify the method for specifying the recognition mode and to make a smoother change.

  Moreover, instead of outputting feedback information each time a word string of a speech recognition result is output, it may be controlled whether or not the feedback generation means 13 generates feedback information. That is, it is possible to determine whether or not the feedback information is generated by determining the error level of the word string of the speech recognition result.

  For example, as shown in FIG. 4, the speech recognition apparatus 10 </ b> A includes an erroneous recognition determination unit 15 that determines an error degree of a word string of a speech recognition result supplied from the speech recognition unit 12. Control information indicating whether or not to generate feedback information may be output to the feedback generation unit 13 according to the degree of occurrence of an error in the word string of the speech recognition result that is output as a result. FIG. 4 is a block diagram showing another example of the overall configuration of the speech recognition apparatus according to the present invention, and an example of the configuration of the speech recognition apparatus 10A in which erroneous recognition determination means 15 is added to the overall configuration of FIG. Is shown.

  In the erroneous recognition determination means 15, for example, when the error level of the word string of the speech recognition result is greater than a predetermined threshold, it is determined that there is a high possibility of recognition mode designation error, and the feedback generation means 13 Control information for instructing the generation of feedback information is output, and conversely, if the error degree of the word string of the speech recognition result is less than or equal to the threshold, it is determined that there are few errors, and feedback generation is performed. Control information that instructs the means 13 that it is not necessary to generate feedback information is output.

  Here, the error degree of the word string of the speech recognition result can use, for example, a scale such as a certainty factor / reliability calculated in the speech recognition process. In addition, when the feedback generation unit 13 inquires of the user whether there is an error in the word sequence of the speech recognition result without using the erroneous recognition determination unit 15, and there is a reaction from the user that the word sequence is incorrect However, the feedback information may be generated only.

  That is, as shown in FIG. 5, the feedback generation unit 13 determines that the voice recognition result is incorrect depending on whether or not the user who has received the inquiry from the feedback generation unit 13 presses the misrecognition button (or mode designation button). May be received to determine whether or not feedback information is generated and control whether to generate feedback information. FIG. 5 is a block configuration diagram showing still another example of the overall configuration of the speech recognition apparatus according to the present invention, in which the user is inquired whether there is an error in the speech recognition result, and erroneous recognition occurrence information returned from the user is shown. 3 shows an example of the configuration of a speech recognition apparatus 10B provided with feedback generation means 13A having a function of determining whether or not feedback information is generated according to.

  As described above, by making an inquiry to the user, it is possible to output feedback information only when there is a high possibility of recognition mode designation error. If feedback information for inquiring whether or not to perform voice recognition processing again is not generated, it is very rare to perform voice recognition again after changing the recognition mode for the same input voice. It does not have to be stored as audio data in the storage means.

  Further, based on whether or not the signal-to-noise ratio (S / N) of the input voice is equal to or greater than a predetermined threshold, it is determined whether or not the input voice is appropriate for performing voice recognition processing. It may be configured to control whether to generate feedback information that inquires whether or not to perform recognition processing again after changing the recognition mode. In other words, if it is determined that there is a high possibility that a correct recognition result is obtained when the input speech changes the recognition mode and the speech recognition process is performed again, the feedback information may be generated and the correct recognition result may be obtained. When it is determined that the value is low, control may be performed so that feedback information is not generated.

  For example, as shown in FIG. 6, the speech recognition apparatus 10 </ b> C includes a situation detection unit 16 that determines whether or not the situation is likely to cause a speech recognition error. Whether or not it is large, whether or not sudden noise is mixed, whether or not the user's volume is low, receives voice information, determines whether there is a high possibility of causing a voice recognition error, and based on the determination result, Control information indicating whether or not to generate feedback information may be output to the feedback generation unit 13. As described above, the status detection unit 16 determines whether or not the input voice is easy to perform voice recognition processing based on the information representing the ease of voice recognition regarding the input voice. Information representing the ease of speech recognition regarding the input speech includes the signal-to-noise ratio (S / N) of the input speech, the magnitude of background noise, the presence or absence of sudden noise, the volume, and the like. FIG. 6 is a block diagram showing a further different example of the overall configuration of the speech recognition apparatus according to the present invention. The status detection unit 16 determines whether or not the input speech is easy to perform speech recognition processing. 10 shows a configuration example of a speech recognition apparatus 10C provided with

  If feedback information for inquiring whether or not to perform recognition processing again is not generated, it is very rare to perform recognition again after changing the recognition mode for the same input speech, so the input speech is stored as speech. The means need not be stored as audio data.

  In addition, prior to the timing when the feedback information is output to the user, the speech information is changed to a recognition mode (second recognition mode) different from the recognition mode (first recognition mode) used for the speech recognition processing. Voice recognition processing may be performed again on the voice data stored in the storage means. In this case, the word string resulting from performing the speech recognition process again based on the second recognition mode is output to the user and, for example, “the recognition mode is changed again to the female recognition mode”. It is sufficient to generate and output feedback information.

The configuration of the preferred embodiment of the present invention has been described above. However, it should be noted that such examples are merely illustrative of the invention and do not limit the invention in any way. Those skilled in the art will readily understand that various modifications and changes can be made according to a specific application without departing from the gist of the present invention. For example, the embodiment of the present invention can be expressed as the following configurations in addition to the configurations (1) and (12) in the means for solving the problems. The numbers (2)-(11) and (13)-(24) below correspond to the item numbers in the claims.
(2) Voice storage means for storing voice data is provided, and the voice recognition processing means stores all or part of the input voice for voice recognition as voice data in the voice storage means, and the feedback generation means The speech recognition apparatus according to (1), wherein the feedback information to be generated includes information for inquiring a user whether to perform speech recognition processing on speech data relating to the input speech stored in the speech storage means.
(3) When a response indicating that voice recognition processing is to be performed again from the user who has received the feedback information is detected, the voice recognition unit performs the voice storage unit based on a condition defined by the second recognition mode. The speech recognition apparatus according to (2), wherein speech recognition processing is performed on speech data relating to the input speech stored in the above.
(4) Prior to detecting a response to perform voice recognition processing again from the user who has received the feedback information, the voice recognition means, based on a condition defined by the second recognition mode, The speech recognition apparatus according to (2), wherein speech recognition processing is performed on speech data relating to the input speech stored in speech storage means.
(5) The voice recognition device according to any one of (2) to (4), wherein the voice data related to the input voice stored in the voice storage means is voice data of a section determined as a voice section of the input voice. .
(6) When there is an instruction from the user who has received the feedback information to change the first recognition mode to another recognition mode as the second recognition mode, the feedback generation means Any one of the above (1) to (5) that changes the recognition mode to a different recognition mode and generates feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice Voice recognition device.
(7) The speech recognition device according to any one of (1) to (6), wherein the feedback information generated by the feedback generation unit includes a recognition mode candidate to be changed in the second recognition mode. .
(8) The feedback generation means controls to generate the feedback information when it is determined that the degree of error in the voice recognition result by the voice recognition means is greater than a predetermined threshold. 7) The voice recognition device according to any one of the above.
(9) The feedback generation unit inquires of the user whether there is an error in the voice recognition result by the voice recognition unit, and generates the feedback information when there is a response from the user that there is an error. The voice recognition device according to any one of (1) to (8), which is controlled as described above.
(10) When there is means for detecting information related to the ease of voice recognition by the voice recognition means, and when the detection means determines that the possibility of an error in voice recognition is lower than a predetermined value based on the information The voice recognition device according to any one of (1) to (9), wherein the feedback generation means generates the feedback information.
(11) The information on the ease of speech recognition is at least one of a background noise level or sudden noise in the input speech, or a signal-to-noise ratio or volume of the input speech (10) Voice recognition device.
(12) The user selects at least one or more recognition modes for performing speech recognition on the input speech from a plurality of recognition modes provided in advance and designates the first recognition mode as the first recognition mode. A speech recognition method for performing speech recognition on the input speech based on conditions defined by a recognition mode and outputting a speech recognition result to the user, which is different from the first recognition mode designated by the user. Change to the second recognition mode, generate feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice, and output the feedback information to the user Voice recognition method.
(13) All or a part of the input voice to be subjected to voice recognition is stored as voice data, and the feedback information is inquired of the user as to whether or not voice recognition processing is performed on the stored voice data related to the input voice. (12) The speech recognition method according to (12), which includes information.
(14) When a response indicating that voice recognition processing is to be performed again is detected from the user who has received the feedback information, the stored voice data related to the input voice is determined based on the condition defined by the second recognition mode. (13) The speech recognition method according to (13) above, wherein speech recognition processing is performed.
(15) Prior to detecting a response to perform voice recognition processing again from the user who has received the feedback information, the voice related to the stored input voice based on the condition defined by the second recognition mode The voice recognition method according to (13), wherein voice recognition processing is performed on data.
(16) The voice recognition method according to any one of (13) to (15), wherein the voice data related to the input voice to be stored is voice data of a section determined to be a voice section of the input voice.
(17) When there is an instruction from the user who has received the feedback information to change the first recognition mode to another recognition mode as the second recognition mode, the recognition is further different from the second recognition mode. The voice recognition method according to any one of (12) to (16), wherein the feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice is generated by changing to the mode.
(18) The speech recognition method according to any one of (12) to (17), wherein the feedback information includes a recognition mode candidate to be changed in the second recognition mode.
(19) The speech recognition method according to any one of (12) to (18), wherein control is performed to generate the feedback information when it is determined that the degree of error in the speech recognition result is greater than a predetermined threshold.
(20) The user is inquired whether there is an error in the speech recognition result, and controls to generate the feedback information when there is a response from the user that there is an error. The voice recognition method according to any one of 19).
(21) The information on the ease of speech recognition by the speech recognition means is detected, and the feedback information is generated when it is determined that the possibility of the speech recognition error is lower than a predetermined value based on the information. The speech recognition method according to any one of (12) to (20).
(22) The information regarding the ease of speech recognition is at least one of a background noise level or presence / absence of sudden noise in the input speech or a signal-to-noise ratio or volume of the input speech. Voice recognition method.
(23) A speech recognition program that implements the speech recognition method according to any one of (12) to (22) as a program executable by a computer.
(24) A program recording medium in which the voice recognition program of (23) is recorded on a computer-readable recording medium.

10 speech recognition device 10A speech recognition device 10B speech recognition device 10C speech recognition device 11 recognition mode setting means 12 speech recognition means 13 feedback generation means 13A feedback generation means 14 output means 15 erroneous recognition determination means 16 situation detection means 20 input means

Claims (24)

  The user selects at least one or more recognition modes for performing speech recognition on the input speech from a plurality of recognition modes provided in advance and designates them as a first recognition mode, and the first recognition mode is In the speech recognition apparatus provided with speech recognition means for performing speech recognition on the input speech based on specified conditions and outputting the speech recognition result to the user, the first recognition mode designated by the user is By changing to a different second recognition mode, feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice is generated at a predetermined timing to the user. A speech recognition apparatus comprising feedback generation means for outputting.   Voice storage means for storing voice data, wherein the voice recognition processing means stores all or part of the input voice for voice recognition as voice data in the voice storage means, and the feedback generation means generates the voice data; The voice recognition according to claim 1, wherein the feedback information includes information for inquiring a user whether or not to perform voice recognition processing on voice data related to the input voice stored in the voice storage means. apparatus.   When a response indicating that voice recognition processing is to be performed again from the user who has received the feedback information is detected, the voice recognition unit stores the response in the voice storage unit based on a condition defined by the second recognition mode. The voice recognition apparatus according to claim 2, wherein voice recognition processing is performed on voice data related to the input voice.   Prior to detecting a response to perform voice recognition processing again from the user who has received the feedback information, the voice recognition means performs the voice storage means based on a condition defined by the second recognition mode. The voice recognition apparatus according to claim 2, wherein voice recognition processing is performed on voice data related to the input voice stored in the voice data. The voice recognition according to any one of claims 2 to 4, wherein the voice data relating to the input voice stored in the voice storage means is voice data of a section determined as a voice section of the input voice. apparatus.   When there is an instruction from the user who has received the feedback information to change the first recognition mode to another recognition mode as the second recognition mode, the feedback generation unit starts from the second recognition mode. 6. The feedback information for changing to a different recognition mode and inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice is generated. The speech recognition apparatus described in 1.   The speech recognition according to any one of claims 1 to 6, wherein the feedback information generated by the feedback generation means includes a recognition mode candidate to be a target of the second recognition mode to be changed. apparatus.   2. The control unit according to claim 1, wherein the feedback generation unit performs control so as to generate the feedback information when it is determined that the degree of error in the speech recognition result by the speech recognition unit is greater than a predetermined threshold. 8. The speech recognition device according to any one of items 7.   The feedback generation means inquires of the user whether there is an error in the voice recognition result by the voice recognition means, and controls to generate the feedback information when there is a response from the user that there is an error. The speech recognition apparatus according to claim 1, wherein   Means for detecting information relating to the ease of speech recognition by the speech recognition means, and when the detection means determines that the possibility of an error in speech recognition is lower than a predetermined value based on the information, the feedback The speech recognition apparatus according to claim 1, wherein the generation unit generates the feedback information.   The information relating to the ease of speech recognition is at least one of a background noise level or presence / absence of sudden noise in the input speech, or a signal-to-noise ratio or volume of the input speech. The speech recognition apparatus according to 10.   The user selects at least one or more recognition modes for performing speech recognition on the input speech from a plurality of recognition modes provided in advance and designates them as a first recognition mode, and the first recognition mode is A speech recognition method for performing speech recognition on the input speech based on a prescribed condition and outputting a speech recognition result to the user, wherein the second recognition mode is different from the first recognition mode designated by the user. Changing to the recognition mode, generating feedback information for inquiring the user whether or not to perform voice recognition again on the same voice data as the input voice, and outputting the feedback information to the user A feature of speech recognition.   All or part of the input speech for speech recognition is stored as speech data, and the feedback information includes information for inquiring the user whether speech recognition processing is to be performed on the stored speech data related to the input speech. The speech recognition method according to claim 12, wherein:   When a response indicating that voice recognition processing is to be performed again is detected from the user who has received the feedback information, voice recognition is performed on the stored voice data related to the input voice based on the condition defined by the second recognition mode. The speech recognition method according to claim 13, wherein processing is performed.   Prior to detecting a response to perform voice recognition processing again from the user who has received the feedback information, the stored voice data related to the input voice is determined based on the conditions defined by the second recognition mode. The voice recognition method according to claim 13, wherein voice recognition processing is performed.   The voice recognition method according to claim 13, wherein the voice data relating to the input voice to be stored is voice data of a section determined as a voice section of the input voice.   When there is an instruction from the user who has received the feedback information to change the first recognition mode to another recognition mode as the second recognition mode, the second recognition mode is changed to a different recognition mode. The voice recognition method according to claim 12, further comprising: generating feedback information for inquiring a user whether to perform voice recognition again on the same voice data as the input voice. .   18. The speech recognition method according to claim 12, wherein the feedback information includes a recognition mode candidate to be a target of the second recognition mode to be changed.   19. The speech recognition method according to claim 12, wherein control is performed so as to generate the feedback information when it is determined that an error degree of the speech recognition result is greater than a predetermined threshold value. .   13. The control unit is configured to inquire a user whether there is an error in the speech recognition result and to generate the feedback information when there is a response from the user that there is an error. The speech recognition method according to any one of 19.   Detecting information related to the ease of speech recognition by the speech recognition means, and generating the feedback information when it is determined that the possibility of an error in speech recognition is lower than a predetermined value based on the information. The speech recognition method according to any one of claims 12 to 20.   The information relating to the ease of speech recognition is at least one of a background noise level or presence / absence of sudden noise in the input speech, or a signal-to-noise ratio or volume of the input speech. The speech recognition method according to 21.   23. A speech recognition program, wherein the speech recognition method according to claim 12 is implemented as a program executable by a computer.   24. A program recording medium, wherein the voice recognition program according to claim 23 is recorded on a computer-readable recording medium.

Images