JP2014149457A - Voice recognition device, electronic apparatus, and control program of voice recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To increase accuracy of voice recognition without losing user's convenience.SOLUTION: A recognition control unit (10) includes: an apparatus condition acquisition unit (11) for acquiring information which shows a condition of a digital camera (100); a candidate phrase determination unit (12) for determining a phrase associated with the information as a candidate phrase; a voice acquisition unit (13) for acquiring voice data; a specific phrase detection unit (14) for detecting a specific phrase from the voice data; and a recognition phrase determination unit (15) for determining the fact that the specific phrase is any of the candidate phrases and determining the phrase as a recognition phrase.

Description

  The present invention relates to a speech recognition device for recognizing a phrase included in speech and an electronic apparatus including the speech recognition device.

  Various speech recognition methods for recognizing words contained in speech are known. In the speech recognition process, various techniques are disclosed for improving the accuracy of the speech recognition.

  For example, in Patent Document 1, a vocabulary with a low frequency of use is set up in a timely manner so that a storage device for speech recognition has a hierarchical structure and a higher phrase is input timely so that a long phrase can be input as the user's proficiency level increases. A technique for improving the accuracy of speech recognition by replacing with a new vocabulary is disclosed. Patent Document 2 discloses a technique for improving the accuracy of speech recognition by presenting words and phrases that are candidates for speech recognition to the user and narrowing down or correcting the recognition results. Further, Patent Documents 3 and 4 improve the speech recognition accuracy by specifying a category including a phrase that is a recognition candidate from speech and presenting the category to the user, thereby allowing the user to narrow down the phrase. Techniques for making them disclosed are disclosed.

JP 2004-325704 A (published on November 18, 2004) JP 2012-022251 A (released February 2, 2012) JP 2001-109492 A (published April 20, 2001) JP 2006-184670 A (published July 13, 2006)

  However, in the conventional techniques as described above, although the voice recognition accuracy can be improved, the operability of the user may be impaired. For example, in the technique disclosed in Patent Document 1, the vocabulary used for speech recognition is determined by the proficiency level of the user. For this reason, when a plurality of users share a device using the above technique, the vocabulary may not match the proficiency level of each user. In such a case, erroneous recognition of the vocabulary and associated malfunctions of the device increase, and as a result, the operability of the user may be impaired. In addition, in the technologies disclosed in Patent Documents 2 to 4, since the user needs to perform a predetermined operation after the first utterance before obtaining the final speech recognition result, the user's operation becomes complicated, The operability is impaired.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a speech recognition apparatus and a speech recognition apparatus control program capable of improving recognition accuracy in speech recognition without impairing user operability. It is to be realized.

  In order to solve the above problems, a speech recognition apparatus according to an aspect of the present invention is a speech recognition apparatus that detects a user's utterance as speech and recognizes a phrase included in the speech. Device status acquisition means for acquiring information indicating the state of the target electronic device, and candidate phrases that are associated with the information indicating the state of the electronic device acquired by the device status acquisition means and that are the target of the speech recognition Candidate phrase determining means for determining the speech, voice data acquiring means for acquiring the user's utterance as voice data, and at least one phrase for specifying the utterance content from the voice data acquired by the voice data acquiring means as a specific phrase The specific word detected by the specific word detection means and the specific word detected by the specific word detection means Identify be any phrase, it is characterized by comprising a recognition word determining means for determining the identified phrase as a recognized word.

  According to one aspect of the present invention, the speech recognition apparatus determines candidate words / phrases to be subjected to speech recognition in accordance with the current state of the electronic device, thereby automatically narrowing down candidate words / phrases to be used for speech recognition. Therefore, it is not necessary for the user to narrow down candidate words, that is, to improve the speech recognition accuracy. Therefore, it is possible to improve the recognition accuracy in voice recognition without impairing the user operability.

It is a block diagram which shows the principal part structure of the digital camera carrying the speech recognition apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the data structure of the phrase table which the said speech recognition apparatus utilizes for speech recognition. It is a flowchart which shows the flow of the speech recognition process performed with the said digital camera. It is a figure which shows the operation | movement outline | summary of the said digital camera. It is a figure which shows the operation | movement outline | summary of the digital camera carrying the speech recognition apparatus which concerns on the 2nd Embodiment of this invention. It is a figure which shows an example of the data structure of the phrase table which the said speech recognition apparatus utilizes for speech recognition.

Embodiment 1
The first embodiment of the present invention will be described as follows. In the present embodiment, an example in which the voice recognition apparatus of the present invention is mounted on a digital camera which is a kind of electronic apparatus will be described. Below, this embodiment is described in detail using FIGS.

[Main part configuration]
First, the main configuration of the digital camera 100 will be described with reference to FIG. FIG. 1 is a block diagram showing a main configuration of the digital camera 100. In the description of the present embodiment, descriptions of portions that are not directly related (for example, a portion connected to an external device or a portion that takes a picture) are omitted. As illustrated, the digital camera 100 (electronic device) includes a recognition control unit 10 as a voice recognition device, a sensor unit 20 (sensor unit, device state specifying unit), a device state specifying unit 21, and a voice detecting unit 30 ( Voice detection means), a phrase table 40, and a device control unit 50.

  The sensor unit 20 detects that the current state of the digital camera 100 has changed. Here, the change in the state of the digital camera 100 indicates a state in which the digital camera 100 has received an arbitrary control instruction. For example, the digital camera 100 is instructed to take a shooting mode, a display mode, various parameter settings, and the like. The sensor unit 20 detects that a control instruction is transmitted to the digital camera 100 and transmits a detection signal to the device state specifying unit 21 described later. The sensor unit 20 may use any detection method as long as it can detect a change in the state of the digital camera 100. For example, the sensor unit 20 may detect a change in state by receiving a setting signal for setting the shooting mode of the digital camera 100. In this case, since the setting signal also changes as the shooting mode is changed, the setting signal may be transmitted to the device state specifying unit 21 as the detection signal.

  The device state specifying unit 21 specifies the state of the digital camera 100 from the detection signal. Here, the state of the digital camera 100 is the state of the digital camera 100 when the digital camera 100 receives the control instruction. Therefore, the state of the digital camera 100 indicates, for example, a state in which a shooting mode, a display mode, and various parameters are set, that is, a control state. When the device state specifying unit 21 specifies the state of the digital camera 100, the device state specifying unit 21 transmits information indicating the specified state to the device state acquiring unit 11 in the recognition control unit 10. The device state specifying unit 21 may use any specifying method as long as the current state of the digital camera 100 can be specified. For example, when the setting signal is received from the sensor unit 20, the shooting mode set by the setting signal may be specified as the current state of the digital camera 100.

  The voice detection unit 30 detects the speech of the user of the digital camera 100 as voice data. The configuration and form of the voice detection unit 30 are not particularly limited. The voice data detected by the voice detection unit 30 is transmitted to the voice acquisition unit 13 of the recognition control unit 10 described later. In addition, the sensor unit 20, the device state specifying unit 21, and the sound detection unit 30 described above may be incorporated in the digital camera 100 or connected as an external device of the digital camera 100.

  The phrase table 40 is information indicating a phrase for matching with a specific phrase. In the word / phrase table 40, information indicating whether or not the word / phrase is a candidate word / phrase and information indicating a control command of the digital camera 100 are stored in association with the word / phrase. Here, the “candidate word / phrase” refers to a word / phrase (to be a candidate for detecting a word / phrase included in a user's utterance in speech recognition) to be compared with a specific word / phrase in the recognition word / phrase determination unit 15 described later. The detailed data structure of the phrase table 40 will be described in detail later. The phrase table 40 is rewritten by the candidate phrase determination unit 12 and is referred to by the recognition phrase determination unit 15 and the control signal output unit 16.

  The recognition control unit 10 controls the overall voice recognition of the digital camera 100. The recognition control unit 10 is realized by, for example, a CPU (central processing unit). More specifically, the recognition control unit 10 includes a device state acquisition unit 11 (device state acquisition unit), a candidate phrase determination unit 12 (candidate phrase determination unit), a voice acquisition unit 13 (voice data acquisition unit), and a specific phrase detection unit 14. (Specific phrase detection means), recognition phrase determination section 15 (recognition phrase determination means), and control signal output section 16 are included.

  The device state acquisition unit 11 acquires information indicating the state of the digital camera 100 from the device state specifying unit 21 and transmits the acquired information to the candidate phrase determination unit 12 described later.

  The candidate phrase determination unit 12 determines the phrase associated with the information indicating the state of the device received from the device state acquisition unit 11 as the candidate phrase that is the target of the voice recognition. Specifically, the candidate word / phrase determination unit 12 determines which words / phrases stored in a word / phrase table 40 to be described later are used as candidate words / phrases. The candidate phrase determination method performed by the candidate phrase determination unit 12 will be described in detail later.

  The voice acquisition unit 13 acquires voice data from the voice detection unit 30 and transmits the acquired voice data to the specific phrase detection unit 14.

  The specific phrase detection unit 14 detects a specific phrase (hereinafter referred to as “specific phrase”) from the phrases included in the voice data transmitted from the voice acquisition unit 13. Here, the “specific phrase” refers to a phrase indicating the content of the user's voice instruction to the digital camera 100 in order to control the digital camera 100. More specifically, the specific phrase detection unit 14 converts voice data into text data, and detects a specific phrase included in the voice data as text data from a previously registered database (not shown). The detected specific phrase is transmitted to the recognized phrase determination unit 15.

  The specific phrase may be detected in a format that can match the specific phrase with the phrase determined as a candidate phrase among the phrases in the phrase table 40 by the recognition phrase determination unit 15 described later. The format is not particularly limited. For example, the specific phrase may be voice data.

  The recognition word / phrase determination unit 15 collates the specific word / phrase with each candidate word / phrase, and determines the candidate word / phrase that matches the specific word / phrase as a result of the collation as the word / phrase indicated by the specific word / phrase. Hereinafter, as a result of the comparison between the specific word and each candidate word, the candidate word determined to match the specific word is simply described as “recognized word”. More specifically, the recognized phrase determination unit 15 collates the character string of the specific phrase with the character string of the candidate phrase stored in the phrase table 40. As a result of the collation, when the character strings of both coincide with each other at a certain ratio or more, it is determined that the candidate phrase is a recognized phrase.

  The control signal output unit 16 reads out a control command associated with the recognized word / phrase determined by the recognized word / phrase determining unit 15 from the word / phrase table 40, creates a control command for controlling various functions of the digital camera 100, and generates the control command. The control command is output to the device control unit 50 as a control signal. Here, the “control command” is a command for controlling various functions of the digital camera 100.

  The device control unit 50 controls various functions of the digital camera 100. Functions controlled by the device control unit 50 are not particularly limited, but specific examples include basic functions that the digital camera 100 has as a camera, such as photography and timers, various setting functions such as shooting mode settings, and menu screens. Display and operation functions.

[Data structure of phrase table]
Next, the detailed data structure of the phrase table 40 will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of a data structure of the phrase table 40 used by the recognition control unit 10 for speech recognition. Specifically, FIG. 2 shows an example of data in the phrase table 40 when the shooting mode (state) of the digital camera 100 is “outdoor”. In FIG. 2, the phrase table 40 is shown as a data structure in a table format, and the data structure of the phrase table 40 is not intended to be limited to the table format. Hereinafter, the same applies to other figures for explaining the data structure. As shown in the figure, the phrase table 40 has a “phrase” column, a “candidate” column, and a “control command” column. The phrase table 40 has a configuration in which a “candidate” column and a “control instruction” column are associated with a “phrase” column. Therefore, if the information of the “word / phrase” column is determined, the information of the “candidate” column and the “control instruction” column is uniquely determined.

  The “word / phrase” column stores words / phrases used by the recognition control unit 10 for voice recognition.

  The “candidate” column stores information as to whether or not the word / phrase stored in the “word / phrase” column in each row is a candidate word / phrase in speech recognition. The information in the “candidate” column is rewritten by the candidate phrase determination unit 12 in accordance with the device status acquired by the device status acquisition unit 11. In FIG. 2, the phrase stored in the “phrase” column in the row where the “candidate” column is “◯” is a target of collation with a specific phrase. That is, the phrase in the row where the “candidate” column is “◯” is a candidate phrase. On the other hand, words stored in the “word / phrase” column of the row where the “candidate” column is blank are not subjected to the above collation.

  The “control command” column is executed by the digital camera 100 when it is determined in each row of the phrase table 40 that the phrase stored in the “phrase” column matches the specific phrase, that is, is a recognized phrase. Stores information indicating control instructions. Here, the “control command” indicates information necessary for the control signal output unit 16 to create a control command. Information in the “control command” column is read by the control signal output unit 16.

  The format of information stored in each column is not particularly limited. The information in the “candidate” column may be freely changeable by the user. The phrase table 40 may be predetermined in the digital camera 100, but may be stored so that it can be rewritten by the user. For example, the user may be able to add a new phrase to the phrase table 40, or the control command associated with the phrase or information on the “candidate” column may be changed. Furthermore, the phrase table 40 does not necessarily have to store control commands. However, when the control command information is stored in the phrase table 40 or another table and is associated with the above phrase, it is possible to control the digital camera 100 corresponding to the recognized phrase at the time of speech recognition. There is a merit that it becomes.

[Determination of candidate phrases by the candidate phrase determination unit]
Next, determination of candidate words and phrases performed by the candidate word and phrase determination unit 12 will be described in detail with reference to FIG. More specifically, the candidate phrase determination unit 12 switches the value of the “candidate” column of the phrase in the phrase table 40 based on the information indicating the device status received from the device status acquisition unit 11. .

  When the shooting mode of the digital camera 100 is “outdoor”, it is predicted that the user rarely speaks a phrase that changes the white balance setting to the indoor setting. As described above, in each state of the digital camera 100, for a word / phrase that is predicted not to be spoken by the user, the “candidate” column in the word / phrase table 40 is left blank, and is excluded from the candidate word / phrase in this state. Specifically, when the candidate word / phrase determination unit 12 receives information indicating the state of the digital camera 100 from the device state acquisition unit 11, the candidate word / phrase determination unit 12 refers to the word / phrase table 40 and searches for a word / phrase associated with the state in advance. . Then, the candidate phrase determining unit 12 sets only the “candidate” column of the phrase associated with the above state to “◯” (the above phrase is a candidate phrase), and the “candidate” column associated with the other phrase. All information is blank.

  In addition, the candidate phrase determination unit 12 may use a phrase associated with a control command that can be executed in the current state of the digital camera 100 as a candidate phrase instead of narrowing down the candidate phrases by the above-described method. In this case, information indicating a control command is stored in the digital camera 100, and the information indicating the control command is associated with a word / phrase in the word / phrase table 40. Further, the candidate word / phrase determination unit 12 may change the word / phrase table to be referred to according to the state of the digital camera 100 instead of rewriting the information in the “candidate” column as described above. In this case, the digital camera 100 includes a word / phrase table corresponding to each state of the device itself, and the “candidate” column is not rewritten, so the “candidate” column is not necessarily required.

[Process flow]
FIG. 3 is a diagram showing a flow of voice recognition processing performed in the digital camera 100. First, information indicating the state of the digital camera is acquired (S100). Specifically, the device state acquisition unit 11 acquires information indicating the state of the digital camera 100 specified by the device state specifying unit 21. Next, the “candidate” column in the phrase table is rewritten according to the information indicating the state of the digital camera acquired in S100 (S102). Specifically, when the candidate phrase determination unit 12 receives the information indicating the state from the device state acquisition unit 11, the candidate phrase determination unit 12 determines whether the “candidate” column of the phrase table 40 corresponds to the state of the digital camera 100 indicated by the received information. Rewrite information. As a result, it is determined which of the phrases included in the phrase table 40 is the candidate phrase for speech recognition.

  Next, when voice is detected (YES in S104), voice data is acquired from the detected voice (S106). Specifically, the voice detection unit 30 acquires voice data of the detected speech of the user of the digital camera 100 by the voice acquisition unit 13. Subsequently, a specific phrase included in the audio data is detected (S108). Specifically, the specific word / phrase detection unit 14 detects a specific word / phrase to be subjected to voice recognition from the voice data acquired by the voice acquisition unit 13. Next, a recognition word / phrase is determined (S110). Specifically, the recognized word / phrase determination unit 15 determines a candidate word / phrase in the word / phrase table 40 that matches the specific word / phrase detected by the specific word / phrase detection unit 14 as a recognized word / phrase.

  Subsequently, the control signal output unit 16 reads a control command corresponding to the recognized word from the word table (S112), creates a control command from the read control command (S114), and outputs the control command (S116). Finally, the digital camera 100 performs control according to the control command (S118).

[Operation overview]
Next, an outline of the operation of the digital camera 100 equipped with the speech recognition apparatus according to the present embodiment will be described with reference to FIG. 4A and 4B specifically show the operation of the digital camera 100 in which the shooting mode is “outdoor”. Here, candidate phrases are determined using the phrase table 40 shown in FIG. That is, as shown in FIG. 2, the candidate phrase determination unit 12 of the digital camera 100 uses the phrase “white balance sunny” as a candidate phrase, but does not consider the phrase “white balance fluorescent lamp” as a candidate phrase.

  In the above case, it is assumed that the user utters the phrase “white balance clear sky” ((a) in FIG. 4). In this case, the above phrase is a candidate phrase. Therefore, the recognized word / phrase determination unit 15 collates the word with the specific word / phrase. Thereby, it is determined that the word / phrase is a recognized word / phrase, and a control command corresponding to the recognized word / phrase (switching the white balance to the setting of “clear sky”) is executed as a control command. That is, the user's voice is recognized and the white balance is changed to “clear sky”. On the other hand, when the user utters the phrase “white balance fluorescent lamp” ((b) in FIG. 4), the recognized phrase determination unit 15 does not collate with the specific phrase because the phrase is not a candidate phrase. That is, the user's voice is not recognized.

  As described above, the digital camera 100 equipped with the voice recognition device according to the present embodiment can exclude words indicating instructions that the user is not supposed to give in the shooting mode from the voice recognition candidates. it can.

  In the first embodiment, an example in which candidate words / phrases to be recognized for speech recognition are determined according to the state of the digital camera 100 is described. However, the present invention is not limited to this, and information indicating the state of the digital camera 100 is used. You may make it determine whether the said phrase shall be made into the said candidate phrase from the character count of the matched phrase. In this case, in a state where the detection accuracy of a specific word / phrase is considered to be lowered, a word / phrase that is unlikely to be erroneously recognized can be set as a candidate word / phrase. In Embodiment 2 below, particularly when the state of the digital camera 100 is a predetermined state (a state where misrecognition is likely to occur), the number of characters in the phrase associated with the information indicating the state of the digital camera 100 is An example will be described in which words and phrases that are more likely to cause erroneous detection of recognized words are determined as the candidate words.

<< Embodiment 2 >>
The second embodiment of the present invention will be described as follows. In the present embodiment, as in the first embodiment, an example in which the speech recognition apparatus of the present invention is mounted on a digital camera which is a kind of electronic apparatus will be described. For convenience of explanation, members having the same functions as those described in the first embodiment are denoted by the same reference numerals and description thereof is omitted.

  The present embodiment is different from the first embodiment in the process in which the candidate phrase determining unit 12 in the recognition control unit 10 shown in FIG. 1 determines the candidate phrase. In the speech recognition process according to the present embodiment, words that are likely to be erroneously recognized are stored in the phrase table so that the accuracy of speech recognition can be ensured as much as possible in a situation where it is predicted that speech cannot be accurately detected (a state in which erroneous recognition is likely to occur). exclude.

  Here, the phrase “prone to misrecognition” is specifically a phrase having a small number of characters (short), a phrase including many characters or words similar to other words, and the like. This is because the phrase as described above has few portions that can be distinguished from other phrases, and is difficult to determine accurately in speech recognition.

  Next, the candidate phrase determination performed by the candidate phrase determination unit 12 in the present embodiment will be described in detail. The candidate word / phrase determination unit 12 according to the present embodiment determines candidate words / phrases by the following method in addition to the determination of candidate words / phrases described in the first embodiment.

  5A to 5D are diagrams showing an outline of the operation of the digital camera 100 in the present embodiment. Here, the digital camera 100 shown in FIGS. 5A and 5B is a mode other than the “self-shooting” shooting mode, and the digital camera 100 shown in FIGS. The shooting mode is “selfie”. FIG. 6 shows a specific example of the phrase table 40 used by the speech recognition apparatus according to the second embodiment of the present invention. More specifically, FIG. 6 shows the phrase table 40 when the shooting mode of the digital camera 100 is “selfie”. Note that the data structure of the phrase table 40 is the same as that in the first embodiment.

  By the way, when the shooting mode (state) of the digital camera 100 is “self-portrait” ((c) and (d) in FIG. 5), the distance between the user who speaks and the voice detection unit 30 of the digital camera 100 is usually This is considered to be farther from the case ((a) and (b) of FIG. 5).

  In the above case, it is considered that the quality of the voice data detected by the voice detection unit 30 is lowered and the detection accuracy of the specific phrase detected based on the voice data is also lowered. As described above, if a word / phrase having a small number of characters is also a candidate word / phrase when the detection accuracy of the specific word / phrase is low, the erroneous detection of the recognized word / phrase increases and the operability of the user is impaired.

  Therefore, as shown in FIG. 6, the candidate phrase determination unit 12 in the present embodiment sets a corresponding “candidate” column for words and phrases such as “photographing”, “menu”, and “browsing” that are less than a predetermined threshold. Leave blank and exclude from candidate words. If the threshold is set to the number of characters that are likely to cause erroneous detection of a recognized word / phrase in speech recognition, candidate words / phrases that are likely to be erroneously detected can be excluded. Therefore, it is possible to reliably prevent erroneous detection of a recognized word / phrase in speech recognition.

  Note that the speech recognition processing in the present embodiment can be further improved in accuracy of speech recognition by being used together with the speech recognition processing in the first embodiment. For example, also in this embodiment, the “candidate” column corresponding to the phrase “timer set” that the user will not utter when the shooting mode is “selfie” may be blank (FIG. 6).

[Operation overview]
Next, an outline of the operation of the digital camera 100 according to the present embodiment will be described with reference to FIG. When the shooting mode of the digital camera 100 is other than “self-shooting” ((a) and (b) in FIG. 5), the candidate word / phrase determining unit 12 of the digital camera 100 selects either “shutter shooting” or “shooting”. The phrase is also a candidate phrase. On the other hand, when the shooting mode is “self-portrait” ((c) and (d) in FIG. 5), “shutter shooting” is a candidate word, but “shooting” is not a candidate word. Therefore, when the shooting mode of the digital camera 100 is other than “self-portrait”, even if the user utters the phrase “shutter shoot” (FIG. 5A) or the phrase “shoot” (FIG. 5) 5 (b)), the user's voice is recognized as a voice and a picture is taken. On the other hand, when the shooting mode of the digital camera 100 is “selfie”, when the user utters the phrase “shutter shooting” ((c) in FIG. 5), the user's voice is recognized. When the phrase “photographing” is issued ((d) in FIG. 5), voice recognition is not performed.

  Thus, in this embodiment, when it is predicted that the quality of the audio data detected by the audio detection unit 30 will be reduced, words that are likely to be erroneously recognized are further excluded from the candidate words, such as “shooting”. As a result, it is possible to achieve both user operability and improvement in voice recognition accuracy.

  In both of the first and second embodiments, the device state acquisition unit 11 acquires information indicating the control state of the digital camera 100 as the device state. However, the present invention is not limited to this, and the environment of the digital camera 100 is not limited thereto. Information indicating the state, that is, detection values such as the brightness around the digital camera 100 and the tilt of the digital camera 100 may be acquired as the device state. In the third embodiment described below, an example in which voice recognition processing is performed using information indicating the environmental state of the digital camera 100 will be described.

<< Embodiment 3 >>
The third embodiment of the present invention will be described as follows. In the present embodiment, as in the first embodiment, an example in which the speech recognition apparatus of the present invention is mounted on a digital camera which is a kind of electronic apparatus will be described. For convenience of explanation, members having the same functions as those described in the first embodiment are denoted by the same reference numerals and description thereof is omitted.

  In the digital camera 100 according to the present embodiment, the sensor unit 20 illustrated in FIG. 1 serves as various sensors that detect information indicating the environmental state of the digital camera 100, that is, detection values such as brightness and inclination. Examples of the sensor and the information include the external or internal temperature of the device detected by the temperature sensor, the tilt of the electronic device detected by the tilt sensor, and the intensity of external light detected by the optical sensor.

  As described above, when the information indicating the environmental state of the digital camera 100, that is, the detection value such as brightness and inclination is acquired as the device state, the candidate phrase determination unit 12 determines the candidate phrase without any user operation. Can be done. That is, instead of determining candidate words / phrases triggered by a user operation such as a change in shooting mode, the candidate words / phrases are automatically determined in accordance with the environment inside or outside the digital camera 100. be able to. Thereby, the accuracy of voice recognition can be improved while improving the operability for the user.

[Modification]
The speech recognition apparatus according to the present invention is not limited to the digital camera described above. The speech recognition apparatus according to the present invention can be installed in any electronic device and perform speech recognition as long as the device can acquire the state of the speech and the electronic device. For example, the voice recognition device may be mounted on a television. In this case, only the channel currently being broadcast is set as a recognition target. The voice recognition device may be mounted on an HDD (hard disc drive) recorder. In this case, only recorded programs in the program data are recognized.

  Further, the voice recognition device is not necessarily built in the electronic device. For example, the state of the electronic device and audio data may be acquired by communicating with the electronic device. For example, the recognition control unit 10 and the voice detection unit 30 may be mounted on a smartphone or the like, and the sensor unit 20 and the device state specifying unit 21 may be mounted on a home appliance such as a television.

[Example of software implementation]
The control blocks (particularly the candidate word determination unit 12 and the recognition word determination unit 15) of the recognition control unit 10 may be realized by a logic circuit (hardware) formed in an integrated circuit (IC chip) or the like, or a CPU ( It may be realized by software using a Central Processing Unit. In the latter case, the recognition control unit 10 includes a CPU that executes instructions of a program that is software that implements each function, and a ROM (Read Only Memory) in which the program and various data are recorded so as to be readable by a computer (or CPU). Alternatively, a storage device (these are referred to as “recording media”), a RAM (Random Access Memory) for expanding the program, and the like are provided. And the objective of this invention is achieved when a computer (or CPU) reads the said program from the said recording medium and runs it. As the recording medium, a “non-temporary tangible medium” such as a tape, a disk, a card, a semiconductor memory, a programmable logic circuit, or the like can be used. The program may be supplied to the computer via an arbitrary transmission medium (such as a communication network or a broadcast wave) that can transmit the program. The present invention can also be realized in the form of a data signal embedded in a carrier wave in which the program is embodied by electronic transmission.

[Summary]
The speech recognition device (recognition control unit 10) according to aspect 1 of the present invention is a speech recognition device that detects a user's speech as speech and recognizes words included in the speech, and is a target of speech operation. Corresponding to device status acquisition means (device status acquisition unit 11) for acquiring information indicating the status of the electronic device (digital camera 100), and information indicating the status of the electronic device acquired by the device status acquisition means, Candidate phrase determination means (candidate phrase determination section 12) for determining candidate phrases to be subjected to voice recognition, voice data acquisition means (voice acquisition section 13) for acquiring the user's speech as voice data, and the voice data Specific phrase detection means (specific phrase detection unit 14) that detects at least one phrase that specifies the utterance content from the voice data acquired by the acquisition means as a specific phrase The specific phrase detected by the specific phrase detection unit is identified as one of the candidate phrases determined by the candidate phrase determination unit, and the identified phrase is determined as a recognition phrase (Recognition word determining unit 15).

  According to the said structure, the candidate word / phrase used as the object of speech recognition can be determined according to the present state of an electronic device. Thus, the electronic device can automatically narrow down candidate words and phrases without requiring an operation from the user. Therefore, it is possible to improve the recognition accuracy in voice recognition without impairing the user operability.

  The speech recognition apparatus according to aspect 2 of the present invention is the speech recognition apparatus according to aspect 1, wherein the candidate phrase determination unit determines the phrase from the number of characters in the phrase associated with the information indicating the state of the electronic device as the candidate phrase. You may decide whether or not.

  According to the above configuration, the candidate word / phrase is determined based on the number of characters of the word / phrase associated with the information indicating the state of the electronic device, so that the recognized word / phrase determining unit calculates the number of characters of the specific word / phrase and the number of characters of the candidate word / phrase. Since the recognition word / phrase is determined as a parameter for determining the recognition word / phrase, the determination accuracy of the recognition word / phrase can be improved. Thereby, it is possible to prevent the user from requesting re-operation due to malfunction of the electronic device due to erroneous voice recognition. Therefore, it is possible to reduce the decrease in user operability and improve the accuracy of voice recognition.

  The speech recognition apparatus according to aspect 3 of the present invention is the speech recognition apparatus according to aspect 2, wherein the candidate phrase determining means further determines that the number of characters in the phrase associated with the information indicating the state of the electronic device is the recognition phrase in speech recognition. Words that are more than the number of characters that are likely to be erroneously detected may be determined as the candidate words.

  According to the above configuration, the candidate word determination unit can exclude words and phrases that are likely to be erroneously recognized and whose number of characters is smaller than a predetermined value from the candidate words. That is, there are few portions that can be distinguished from other words and phrases, and words that are difficult to be accurately identified in speech recognition can be excluded from candidate phrases. Therefore, the operability of the user and the accuracy of voice recognition can be further improved.

  The speech recognition apparatus according to Aspect 4 of the present invention is the sensor unit according to any one of Aspects 1 to 3, wherein the apparatus state acquisition unit measures apparatus information (environment information) indicating an external or internal environment of the electronic apparatus. The device information may be acquired from the (sensor unit 20) as the state of the electronic device.

  According to the said structure, a candidate word / phrase can be determined based on the apparatus information which the sensor part acquired. Therefore, it is possible to automatically determine candidate words in accordance with the environment inside or outside the electronic device. Thereby, the accuracy of voice recognition can be improved while improving the operability for the user.

  An electronic apparatus according to an aspect 5 of the present invention is an electronic apparatus including the speech recognition device according to any one of the above aspects 1 to 4, and a voice detection unit (voice detection unit 30) that detects the user's utterance; Device state specifying means (device state detection unit 20) for specifying the state of the device itself.

  According to the above configuration, the electronic device determines a candidate word / phrase from the state of the electronic device specified by the device state specifying unit, and performs voice recognition using the user's voice detected by the voice detecting unit and the candidate word / phrase. It can be carried out. Accordingly, the electronic device can perform speech recognition using an appropriate word / phrase as a candidate word / phrase in the current state of the own device.

  The speech recognition apparatus according to each aspect of the present invention may be realized by a computer. In this case, the speech recognition apparatus is realized by the computer by operating the computer as each unit included in the speech recognition apparatus. A control program for the speech recognition apparatus and a computer-readable recording medium on which the control program is recorded also fall within the scope of the present invention.

  The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention. Furthermore, a new technical feature can be formed by combining the technical means disclosed in each embodiment.

  The present invention is suitable for an electronic device that can be operated by voice recognition.

  DESCRIPTION OF SYMBOLS 10 recognition control part (voice recognition apparatus), 11 apparatus state acquisition part (apparatus state acquisition means), 12 candidate phrase determination part (candidate phrase determination means), 13 voice acquisition part (voice acquisition means), 14 specific phrase detection part ( Specific phrase detection unit), 15 recognition phrase determination unit (recognition phrase determination unit), 20 sensor unit (sensor unit), 21 device state identification unit (device state identification unit), 30 voice detection unit (voice detection unit), 100 digital Camera (electronic equipment)

Claims (5)

A speech recognition device that detects a user's utterance as speech and recognizes speech included in the speech,
Device status acquisition means for acquiring information indicating the status of an electronic device to be subjected to voice operation;
Candidate phrase determination means for determining candidate phrases that are associated with the information indicating the state of the electronic device acquired by the apparatus status acquisition means and that are the target of speech recognition;
Voice data acquisition means for acquiring the user's speech as voice data;
Specific phrase detection means for detecting at least one phrase that specifies utterance content from the voice data acquired by the voice data acquisition means as a specific phrase;
A recognized word / phrase determining means for specifying that the specific word / phrase detected by the specific word / phrase detecting means is one of the candidate words / phrases determined by the candidate word / phrase determining means, and determining the identified word / phrase as a recognized word / phrase; ,
A speech recognition apparatus comprising:   2. The voice according to claim 1, wherein the candidate phrase determination unit determines whether or not to set the phrase as the candidate phrase from the number of characters of the phrase associated with the information indicating the state of the electronic device. Recognition device.   The device status acquisition means acquires the device information as information indicating the status of the electronic device from a sensor unit that measures device information indicating the external or internal environment of the electronic device. The speech recognition apparatus according to 2. An electronic device comprising the voice recognition device according to any one of claims 1 to 3,
Voice detection means for detecting the user's utterance;
An electronic device comprising: device state specifying means for specifying the state of the device itself.   A control program for operating the speech recognition apparatus according to any one of claims 1 to 3, wherein the control program causes a computer to function as each of the means.

Images