JP5336788B2 - Speech recognition apparatus and program - Google Patents

Description

  The present invention relates to a voice recognition technique, and more particularly, to a technique for notifying a user that a word is out of a vocabulary when a voice signal input together with a voice recognition request from the user is out of the vocabulary.

2. Description of the Related Art Conventionally, a technology for registering words in a language model from a document file described in a free format that includes words to be added to a language model (speech recognition dictionary) in a speech recognition apparatus is known. The speech recognition apparatus disclosed in Patent Document 1 reads a document file containing unknown words, performs morphological analysis, etc., extracts words, and extracts words that do not exist in the language model as unknown words. The extracted unknown words are displayed in a list, and after the user corrects or deletes the readings and parts of speech assigned to the unknown words and unknown words, the selected unknown words are collectively registered in the language model.
JP 2003-316376 A

  However, in the above prior art, since words are extracted from a document file described in a free format and registered, unnecessary words may be added to the language model. As a result, there is a concern that the language model is enlarged and the speech recognition performance is deteriorated.

  The present invention has been made in view of such circumstances, and when a voice signal input together with a voice recognition request from a user is out of the vocabulary, the user is notified that it is out of the vocabulary. An object of the present invention is to provide a speech recognition apparatus and program capable of performing the above.

  (1) In order to achieve the above object, the present invention takes the following measures. That is, the speech recognition device according to the present invention is a speech recognition device that notifies a user that a vocabulary is out of a vocabulary when a speech signal input together with a speech recognition request from the user is out of the vocabulary. A recognition processing request receiving means for receiving a speech recognition request and an input of a speech signal from, a non-vocabulary determining unit for determining whether or not the input speech signal is out of a vocabulary, an input using an acoustic model and a language model Voice recognition processing means for performing voice recognition processing of the received voice signal, and when the input voice signal is out of the vocabulary, the user is notified that it is out of the vocabulary .

  In this way, if the input speech signal is out of the vocabulary, the user is notified that the vocabulary is out of the vocabulary, so it is possible to prevent useless recurrence and improve the convenience of the speech recognition device. It becomes possible to make it.

  (2) In addition, the speech recognition apparatus of the present invention requests the user to register in the language model speech that is outside the vocabulary when the input speech signal is outside the vocabulary. It is characterized by doing.

  In this way, when the input speech signal is out of the vocabulary, the user is required to register the speech that is considered out of the vocabulary in the language model. It can be increased.

  (3) Moreover, in the speech recognition apparatus of the present invention, the out-of-vocabulary determination unit includes similarity measurement means for measuring the similarity between the speech signal input together with the speech recognition request and the speech signal input in the past. A non-vocabulary utterance determination means for determining that the voice is a recurrent utterance outside the vocabulary when there is a past voice having a high similarity to the voice signal input together with the voice recognition request within a predetermined period from the current time. It is characterized by providing.

  As described above, when a voice signal having a high degree of similarity is input within a predetermined period from the current time, it is possible to make the user avoid the vocabulary utterance by determining that the utterance is out of the vocabulary. It becomes. Note that the number of recurrent voices within a predetermined period is N (N is a natural number).

  (4) The program of the present invention has a function of notifying the user that the vocabulary is out of the vocabulary when the voice signal input together with the voice recognition request from the user is out of the vocabulary. A program for realizing, a process of accepting a voice recognition request and a voice signal input from a user, a process of determining whether or not the inputted voice signal is outside the vocabulary, and the inputted voice signal outside the vocabulary If there is, the process of notifying the user that it is out of the vocabulary, and if the input audio signal is out of the vocabulary, the content of the utterance that is determined to be out of the vocabulary to the user A series of processes including a process for requesting registration in the language model is converted into a command that can be read and executed by a computer.

  In this way, if the input speech signal is out of the vocabulary, the user is notified that the vocabulary is out of the vocabulary, so it is possible to prevent useless recurrence and improve the convenience of the speech recognition device. It becomes possible to make it.

  According to the present invention, when the input speech signal is out of the vocabulary, the user is notified that the vocabulary is out of the vocabulary. Can be improved.

  Next, embodiments according to the present invention will be described with reference to the drawings. FIG. 1 is a diagram illustrating a schematic configuration of a speech recognition apparatus according to the present embodiment. In FIG. 1, a recognition process request receiving unit 10 receives a voice recognition request, voice data, and a user ID from a user terminal, outputs voice data and a user ID to the voice recognition processing unit 20, and performs a recognition process. Instruct. The recognition result is returned to the user terminal. The recognition process request receiving means 10 also includes an out-of-vocabulary notifying means 11 for notifying the user that it is out of the vocabulary, and a dictionary updating means 12 for accepting registration from the user and updating the language model 22. Yes.

  The speech recognition processing means 20 performs recognition processing using the acoustic model 21 and the language model 22 in accordance with the recognition request received from the recognition processing request receiving means 10, and returns the recognition result to the recognition processing request receiving means 10.

  When a voice is input, the out-of-vocabulary determination unit 30 compares the input voice with a past input voice and measures the similarity. If the voices input N times in a short time (N is a natural number) are similar, it is determined that the utterance is outside the vocabulary. If it is determined that it is out of the vocabulary, it instructs the outside vocabulary notifying means 11 to notify the user that it is out of the vocabulary. On the other hand, if it is determined that it is not out of the vocabulary, the routine proceeds to normal speech recognition processing. The non-vocabulary notifying means 11 notifies the user that it is out of the vocabulary and requests registration to the language model (dictionary). When the registration is performed by the user, the extra-vocabulary notification unit 11 receives the registration, and the dictionary update unit 12 registers the word in the language model 22.

  FIG. 2 is a diagram showing the concept of the extra-vocabulary notification means and the dictionary update means. If the non-vocabulary determination unit 30 determines that it is outside the vocabulary, “non-vocabulary flag information” indicating that is input to the non-vocabulary notification means 11 (step S11). Next, the non-vocabulary notification means 11 transmits a “non-vocabulary notification” and a “word registration request (request)” to the user (step S12). When the dictionary updating means 12 receives the “word / reading” information from the user (step S13), it reflects it in the language model (step S14). In this case, for example, in the case of descriptive grammar, it is registered in a dictionary and a grammar dictionary. If it is a dictation, it is added to the N-gram.

  FIG. 3 is a diagram illustrating a functional configuration of the out-of-vocabulary determination unit 30. The out-of-vocabulary determination unit 30 includes a similarity measurement unit 31 and a past history DB 33. When it is determined by the similarity measurement unit 31 that the word is out of the vocabulary, “non-vocabulary flag information” indicating that is input from the similarity measurement unit 31 to the out-of-vocabulary notification unit 11.

  FIG. 4 is a diagram showing a schematic configuration of the past history DB 33 shown in FIG. The stored voice DB 33a is a DB that stores input voice. The voice data to be stored may be spectrum data, cepstrum data, VQ data, etc. in addition to voice data in the PCM format or the like. The recognition result DB 33b is a DB that accumulates recognition results. The recognition results that are accumulated are recognized characters and recognition scores. The recognition score may be further held separately for acoustic likelihood and language probability. The access information DB 33c is a DB that accumulates access information. The access information to be accumulated includes access time, access user ID, voice data name stored in the corresponding stored voice DB, and recognition result file name stored in the corresponding recognition result DB.

FIG. 5 is a block diagram showing a schematic configuration of the similarity measuring means 31 shown in FIG. The similarity measurer 31 determines the similarity between the input voice and accumulated voice data uttered in the past. Further, after the recognition process, the distance of the recognition result (character string) is obtained and stored in the distance information table between recognition results. When the obtained distance between the audio data is equal to or less than the threshold, it is determined that the two are similar. In FIG. 5, the similarity determination control means 31a receives the input voice data and the user ID from the recognition process request receiving means 10, and then acquires the voice data of the same user ID in the accumulated voice DB 33a. After receiving the input voice data and the user ID from the recognition process request receiving means 10, the access information analyzing means 31b acquires the access information of the corresponding user ID and selects the voice data used for determining the similarity. Those satisfying the following conditions are selected as audio data for measuring the similarity.
(Condition) The voice of the same user generated N times (N is a natural number) within T from the current time.

  The data processing unit 31c processes the input voice and the voice data acquired from the stored voice DB 33a into the same type of data format. For example, if both are audio data such as PCM or spectral domain data, they are processed into spectral domain data, cepstrum domain data, VQ data, and the like. If both are cepstrum area data, they are processed into cepstrum area data, VQ data, and the like. When both are VQ data, they are left as they are.

  The distance calculation means 31d calculates the distance between the voice data or the distance of the recognition result. The obtained distance is output to the inter-speech distance information table, the recognition result inter-distance information table, and the extra-vocabulary utterance determination unit 31e.

  The out-of-vocabulary utterance determination unit 31e determines that they are similar when the distance obtained by the distance calculation unit 31d is equal to or less than a threshold value. The determination result is output to the non-vocabulary notification unit 11.

FIG. 6 is a diagram showing the concept of the distance calculation method performed by the distance calculation means 31d shown in FIG. In this distance calculation, the distance between two voices having different numbers of frames is obtained using DTW (Dynamic Time Stretching Method). Examples of distances between frames include the following distance measures.
(1) Logarithmic spectrum, LPC spectrum, cepstrum, Euclidean distance of VQ data.
(2) Maximum likelihood spectral distance using LPC spectrum.
(3) Cosh scale.

  FIG. 7 is a diagram showing the concept of the operation of the out-of-vocabulary utterance determination unit 31e shown in FIG. The out-of-vocabulary utterance determination unit 31e has an input voice that is out of the vocabulary when the access time is within T from the current time and occurs N times or more and the similarity is high (distance is equal to or less than a threshold). Is determined.

  FIG. 8 is a flowchart showing the operation of the speech recognition apparatus according to this embodiment. When receiving the recognition request (step S80), the speech recognition apparatus inputs the input speech data and user ID to the out-of-vocabulary determination unit 30. Then, in the out-of-vocabulary determining unit 30, the similarity measuring unit 31 measures the similarity. Next, the non-vocabulary determination unit 30 determines whether or not it is out of the vocabulary (step S81). If the result of this determination is not outside the vocabulary, normal recognition processing is performed (step S82). That is, voice data and a user ID are input from the recognition process request receiving unit 10 to the voice recognition processing unit 20.

  On the other hand, if it is determined in step S81 that it is out of the vocabulary, the out-of-vocabulary determination unit 30 transmits the out-of-vocabulary flag information to the out-of-vocabulary notification means 11. Then, the non-vocabulary notifying means 11 notifies the user that it is out of the vocabulary (step S83). In addition, a “word registration request (request)” is transmitted to the user (step S83). Next, when “word / reading” is received from the user (step S84), the information is transmitted to the dictionary updating means 12. The dictionary update unit 12 registers words in the language model 22 (step S85).

  The characteristic operations of the present invention as described above can be performed by causing a computer to execute a program. That is, the program of the present invention is a program for realizing a function of notifying a user that a word is out of the vocabulary when the voice signal input together with the voice recognition request from the user is out of the vocabulary. A process for receiving a voice recognition request and input of a voice signal from a user, a process for determining whether the input voice signal is out of the vocabulary, and a case where the input voice signal is out of the vocabulary Is a process of notifying the user that it is out of the vocabulary, and if the input speech signal is out of the vocabulary, the user is sent to the language model of the speech that is out of the vocabulary. A series of processes including a process for requesting registration of a computer is converted into a command that can be read and executed by a computer.

  In this way, if the input speech signal is out of the vocabulary, the user is notified that the vocabulary is out of the vocabulary, so it is possible to prevent useless recurrence and improve the convenience of the speech recognition device. It becomes possible to make it.

It is a figure which shows schematic structure of the speech recognition apparatus which concerns on this embodiment. It is a figure which shows the concept of a non-vocabulary notification means and a dictionary update means. It is a figure which shows the function structure of the non-vocabulary determination part. It is a figure which shows schematic structure of the past log | history DB33 shown in FIG. It is a block diagram which shows schematic structure of the similarity measuring means 31 shown in FIG. It is a figure which shows the concept of the distance calculation method which the distance calculation means 31d shown in FIG. 5 performs. It is a figure which shows the concept of operation | movement of the vocabulary utterance determination means 31e shown in FIG. It is a flowchart which shows operation | movement of the speech recognition apparatus which concerns on this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Recognition process request | requirement receiving means 11 Out-of-vocabulary notification means 12 Dictionary update means 20 Speech recognition processing means 21 Acoustic model 22 Language model 30 Out-of-vocabulary determination part 31 Similarity measurement means 31a Similarity determination control means 31b Access information analysis means 31c Data processing Means 31d Distance calculation means 31e Out-of-vocabulary utterance judgment means 33 Past history DB
33a Accumulated voice DB
33b Recognition result DB
33c Access information DB
33d Recurrence type DB

Claims (2)

A voice recognition device for notifying a user that a voice signal input together with a voice recognition request from the user is out of the vocabulary.
A recognition process request receiving means for receiving an input of a voice recognition request and a voice signal from a user;
An out-of-vocabulary determination unit for determining whether the input audio signal is out of the vocabulary; and
Using an acoustic model and a language model, and a voice recognition processing means for performing speech recognition processing of the input audio signal,
The out-of-vocabulary determination unit
Similarity measuring means for measuring the similarity between the voice signal input together with the voice recognition request and a voice signal input in the past;
An out-of-vocabulary utterance determination unit that determines that a recurrent utterance is out of the vocabulary when there is a past sound having a high similarity to the sound signal input together with the sound recognition request within a predetermined period from the current time. ,
If the input speech signal is out of the vocabulary, the user is notified that the speech is out of the vocabulary, and the registration of the speech that is out of the vocabulary is requested to the language model. Voice recognition device. A program for realizing a function of notifying a user that a voice signal input together with a voice recognition request from a user is out of the vocabulary when the voice signal input from the user is out of the vocabulary,
A process of accepting a voice recognition request and a voice signal input from a user;
The similarity between the voice signal input together with the voice recognition request and the voice signal input in the past are measured, and the degree of similarity between the voice signal input together with the voice recognition request is high within a predetermined period from the current time. When there is a past voice, a process for determining whether the input voice signal is outside the vocabulary by determining that the voice is a recurrent voice outside the vocabulary ; and
If the input audio signal is out of vocabulary, processing to notify the user that it is out of vocabulary; and
When the input speech signal is out of the vocabulary , processing for requesting the user to register the content of the utterance that is out of the vocabulary into the language model used for the speech recognition processing; A program characterized in that a series of processes including are converted into commands that can be read and executed by a computer.