JP6866731B2 - Speech recognition device, speech recognition method, and program - Google Patents

Description

The present invention relates to a voice recognition device, a voice recognition method, and a program, and more particularly to a technique for improving voice recognition accuracy.

Conventionally, a user interface such as a smartphone or a car navigation system that uses voice input has been used. In addition, AI (Artificial Intelligence) and interactive robots have begun to spread, and opportunities for devices and humans to communicate by voice are increasing. Therefore, there is a demand for an accurate voice recognition system. For example, Patent Document 1 describes a voice dialogue device having a function of smoothing a dialogue with a user by estimating the intention and emotion of the user included in the words spoken by the user.

Further, in the conventional voice recognition process, when there are a plurality of conversion candidates as a result of voice recognition, a method in which the system automatically selects the first candidate is common. For example, when the voice "Do you have a recommended sword?" Is input, the voice recognition result is "No. 1: What is the recommended big sword?", "No. 2: What is the recommended experience?" "No. 3: What is the recommended experience?", "No. 4: What is the recommended experience?", "No. 5: What is the recommended big sword?" Conversion candidates are obtained. In this case, if it is a sightseeing scene, No. 2, 3 and 4 are correct conversion candidates, but if you are talking about the game, No. 1 or 5 is the correct conversion. Therefore, in order to perform appropriate audio document conversion, it is necessary to grasp the situation such as the scene and the flow of conversation, and output the candidates accordingly.

Japanese Unexamined Patent Publication No. 2006-313287

For example, the above-mentioned Patent Document 1 describes a technique for estimating and determining the intention (question, question, suspicion) of input voice by using non-verbal information such as voice, image, and physiological signal for the user's emotion and physiological state. It is disclosed. However, since Patent Document 1 estimates the dialogue content only from the user information based on the user's emotions and physiological states, it can correspond to the recognition of the alternative form such as a simple response sentence, but it is ambiguous as described above. Therefore, it is difficult to properly recognize complicated conversation contents.

The present invention has been made in view of such a problem, and provides a voice recognition device, a voice recognition method, and a program capable of obtaining a voice recognition result suitable for a situation such as a conversation scene or a context. The purpose.

In order to solve the above-mentioned problems, the first invention comprises a voice acquisition means for acquiring voice data, a voice recognition means for recognizing the acquired voice data and obtaining a plurality of conversion candidates as a voice recognition result, and a voice data acquisition time. A situation determination means for determining the situation of the above, an extraction means for extracting a conversion candidate suitable for the situation determined by the situation determination means from a plurality of conversion candidates obtained by the voice recognition means , and a word associated with the situation. comprising storage means for storing said status determining means, the word extracted from the speech data, a speech recognition apparatus according to claim and this determines the status based on the word.

According to the first invention, the voice data of the conversation is recognized, a plurality of conversion candidates are obtained as the voice recognition result, the conversation situation (scene or context) is determined, and the conversation situation is selected from the plurality of conversion candidates. Extract suitable conversion candidates. As a result, it becomes possible to obtain a voice recognition result suitable for a conversation situation such as a scene or a context, and it is possible to provide a voice recognition device having high voice recognition accuracy. Since the situation determination means obtains the situation based on the content of the conversation, it is possible to estimate the situation from the content of the conversation itself and extract an appropriate voice recognition result.

Te first invention smell, before Symbol status determination unit may further determine the said status on the basis of the ambient sound to be acquired. As a result, the situation in which the conversation is taking place can be obtained from the characteristics of the surrounding sounds, so that the voice recognition result can be obtained more accurately.

Further, the situation determination means may further obtain the situation based on the acquired image. Further, the situation determination means may further obtain the situation based on the acquired position information. By obtaining the situation of conversation not only from the characteristics of sound but also from images, position information, etc., it is possible to accurately obtain the situation, and thereby it is possible to obtain an accurate voice recognition result.

The second invention is a voice acquisition means for acquiring voice data, a voice recognition means for recognizing the acquired voice data and obtaining a plurality of conversion candidates as a voice recognition result, and a situation determination for determining a situation at the time of voice data acquisition. means and extraction means and the status of the history of the conversation was determined by pre-Symbol status determination means for extracting a conversion candidates for the situations determined by the condition determining means from among a plurality of conversion candidates obtained by the speech recognition means Bei example and a storage means to store, the extraction means, based on the status of the conversation history stored in the storage means, the status of the conversation from the plurality of conversion candidates obtained by the speech recognition means Rukoto to extract appropriate conversion candidate is a speech recognition apparatus according to claim.
According to the second invention, the voice data of the conversation is recognized, a plurality of conversion candidates are obtained as the voice recognition result, the conversation situation (scene or context) is determined, and the conversation situation is selected from the plurality of conversion candidates. Extract suitable conversion candidates. As a result, it becomes possible to obtain a voice recognition result suitable for a conversation situation such as a scene or a context, and it is possible to provide a voice recognition device having high voice recognition accuracy. In addition, it is possible to recognize the flow of the conversation from the history of the conversation situation, and it is possible to obtain a voice recognition result suitable for the context more accurately.

A third invention includes a step in which a computer acquires voice data of a conversation, a step of recognizing the acquired voice data and obtaining a plurality of conversion candidates as a voice recognition result, and a step of determining a conversation situation. The step of determining the situation includes the step of extracting a conversion candidate suitable for the situation of the conversation from a plurality of conversion candidates and the step of memorizing the word associated with the situation, and the step of determining the situation includes the word. extracted from the voice data, a voice recognition method characterized by and this determines the status based on the word.

According to the third invention, the computer recognizes the voice data of the conversation, obtains a plurality of conversion candidates as the voice recognition result, determines the conversation situation, and is suitable for the conversation situation from the plurality of conversion candidates. Extract conversion candidates. As a result, it becomes possible to obtain a voice recognition result suitable for a conversation situation such as a scene or a context, and it is possible to improve the voice recognition accuracy. Since the step of determining the situation obtains the situation based on the content of the conversation, it is possible to estimate the situation from the content of the conversation itself and extract an appropriate voice recognition result.

A fourth invention is a voice acquisition means for acquiring voice data of conversation by using a computer.
The voice recognition means that recognizes the acquired voice data and obtains a plurality of conversion candidates as the voice recognition result, the situation determination means that determines the conversation situation, and the situation determination means from among the plurality of conversion candidates obtained by the voice recognition means. It is a program for functioning as an extraction means for extracting conversion candidates suitable for a conversation situation determined by the above, and a storage means for storing a word associated with the situation, and the situation determination means converts the word into voice data. It is a program characterized by extracting from and determining the situation based on the word.

According to the fourth invention, the computer can function as the voice recognition device of the first invention.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to provide a voice recognition device, a voice recognition method, and a program capable of obtaining a voice recognition result suitable for a situation such as a scene or a context.

Configuration diagram of voice recognition device 1 Configuration diagram of the computer 10 that functions as the voice recognition device 1. A flowchart showing the flow of the voice recognition process executed by the voice recognition device 1. Example of genre data 5 which is data that associates speech recognition conversion candidates with conversation genres Specific examples of situation judgment results based on conversation content and situation judgment results based on sound An example of a voice recognition system 100 using the voice recognition device 1 according to the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a diagram showing a configuration of a voice recognition device 1 according to the present invention. The voice recognition device 1 includes a voice acquisition unit 21, a voice recognition unit 22, a situation determination unit 23, a voice recognition result extraction unit 24, an output unit 25, and a sound analysis unit 26. Further, in addition to these configurations, an image acquisition unit 27, an image analysis unit 28, and a position information acquisition unit 29 may be provided.

FIG. 2 is a diagram showing a configuration example of a computer 10 that functions as a voice recognition device 1. As shown in FIG. 2, the computer 10 includes a control unit 11, a storage unit 12, a media input / output unit 13, a peripheral device I / F unit 14, an input unit 15, a display unit 16, a communication control unit 17, a microphone 18, and the like. It is configured by being connected via a bus 19. When the computer 10 functions as the voice recognition device 1, the control unit 11 of the computer 10 uses each part (voice acquisition unit 21, voice recognition unit 22, situation determination unit 23, voice recognition result extraction) of the voice recognition device 1 shown in FIG. A program describing the functions of the unit 24, the output unit 25, the sound analysis unit 26, the image acquisition unit 27, the image analysis unit 28, and the position information acquisition unit 29) is executed.

The control unit 11 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), and a RAM (Random Access).
Memory) etc.
The CPU calls and executes a program stored in the storage unit 12, ROM, recording medium, etc. in the work memory area on the RAM, and drives and controls each unit connected via the bus 19. The ROM permanently holds a boot program of the computer 10, a program such as a BIOS, data, and the like. The RAM temporarily holds the loaded program and data, and also includes a work area used by the control unit 11 to perform various processes. By reading and executing the above program, the control unit 11 reads and executes each unit of the voice recognition device 1 shown in FIG. 1 (sound acquisition unit 21, voice recognition unit 22, situation determination unit 23, voice recognition result extraction unit 24, output unit). 25, and the sound analysis unit 26, etc.). That is, the control unit 11 acquires the voice data input from the microphone 18 or the communication control unit 17, and executes the voice recognition process (see FIG. 3) described later for the acquired voice data.

The storage unit 12 is, for example, a storage device such as a hard disk drive. The storage unit 12 stores a program executed by the control unit 11, data necessary for executing the program, an operating system, and the like. These program codes are read by the control unit 11 as necessary, transferred to the RAM, read by the CPU, and executed.

The media input / output unit 13 is, for example, a drive device for various recording media (media) such as a CD, DVD, and MO, and inputs / outputs (writes / reads) data to / from the media.

The peripheral device I / F (interface) unit 14 is a port for connecting peripheral devices, and transmits / receives data to / from the peripheral device via the peripheral device I / F unit 14. The peripheral device I / F unit 14 is composed of USB or the like, and usually has a plurality of peripheral device I / Fs. The connection form with peripheral devices may be wired or wireless.

The input unit 15 is, for example, a pointing device such as a keyboard and a mouse, an input device such as a numeric keypad, and outputs the input data to the control unit 11.
The display unit 16 is composed of, for example, a display device such as a liquid crystal panel or a CRT monitor, and a logic circuit (video adapter or the like) for executing display processing in cooperation with the display device, and is input under the control of the control unit 11. Display the display information on the display device. The input unit 15 and the display unit 16 may be a touch panel display in which an input device such as a touch panel is integrally provided on the display screen.

The communication control unit 17 has a communication control device, a communication port, and the like, and controls communication with the network 3 and the like.
The microphone 18 collects voice and inputs it to the control unit 11 as voice data.
The bus 19 is a route that mediates the transfer of control signals, data signals, and the like between the devices.

The functional configuration of the voice recognition device 1 according to the present invention will be described with reference to FIG.
The voice acquisition unit 21 acquires the voice data of the conversation. The voice data of the conversation may be voice data input from the microphone 18 as long as the voice recognition device 1 includes the microphone 18, or is communicated and connected to the voice recognition device 1 via the communication control unit 17 and the network 3. It may be input from the device.

The voice recognition unit 22 voice-recognizes the content of the user's utterance of the acquired voice data, and obtains one or a plurality of conversion candidates as the voice recognition result. The voice recognition unit 22 has voice recognition data such as a speech dictionary, an acoustic model, a language model, etc., which associates speech data with words, and uses these speech recognition data to perform speech sound and Performs voice recognition processing that analyzes the language and converts the content of the utterance into text. The text that is the voice recognition result is output to the conversation content determination unit 231 of the situation determination unit 23.

The situation determination unit 23 determines the situation of the conversation. Specifically, the conversation situation is the content (genre) of the conversation and the surrounding situation (scene) such as the place and purpose where the conversation is taking place. As shown in FIG. 1, the situation determination unit 23 includes a conversation content determination unit 231 for determining the conversation content and a surrounding situation determination unit 232 for determining the surrounding situation.

The conversation content determination unit 231 obtains the content of the conversation by analyzing the words included in the sentence of the conversation. The content of the conversation requested here is the genre or purpose of the conversation. A genre is a classification that indicates what a conversation is about, such as "sightseeing," "game," "food and drink," "movie," "school," "medical care," and so on. The purpose is a classification indicating the purpose of the conversation, such as "customer service" or "chat". For example, the conversation content determination unit 231 stores in advance the data associated with the word and the genre as the genre data 5 (see FIG. 4) in the storage unit 12, and by referring to the genre data 5, the conversation content (see FIG. 4). Genre) is judged. Each word may be included across a plurality of genres. The conversation content determination unit 231 obtains a candidate for conversation content from words included in one or a plurality of sentences obtained as a voice recognition result. Further, it is desirable that not only the sentence to be judged but also the sentence recognized from the voice data input before that (speech recognition result) is included in the judgment target. As a result, the content of the conversation can be obtained in consideration of the context.

The surrounding condition determination unit 232 obtains the surrounding condition such as a place based on the characteristics of the surrounding sound data when the voice data of the conversation is input. The sound analysis unit 26 extracts ambient sound data from the voice data acquired by the voice acquisition unit 21, extracts features of the surrounding sound data, and outputs the characteristics to the ambient condition determination unit 232. The surrounding situation determination unit 232 stores in advance in the storage unit 12 data that associates the characteristics and the situation of the extracted sound data as sound analysis data, and determines the surrounding situation of the conversation based on the sound analysis data. To do. For example, the sound analysis data of a "restaurant" includes features such as the sound of tableware, conversation at a table, and the sound of customer service. In addition, the sound analysis data of the "amusement park" includes features such as cheers and attraction sounds. The ambient condition determination unit 232 obtains one or a plurality of candidates for the ambient condition from the characteristics of the sound data.

The surrounding situation determination unit 232 may determine the surrounding situation based not only on the sound analysis but also on the image and the position information. Specifically, as shown in FIG. 1, an image analysis unit 28 that analyzes an image (video or still image) of a state or place during a conversation by the image acquisition unit 27 is provided, and the image analysis unit 28 talks. Surrounding conditions such as location and purpose may be determined. Further, a position information acquisition unit 29 such as GPS (Global Positioning System) is further provided, and the surrounding condition determination unit 232 obtains a conversation place (store or facility) or the like based on the position information, map data, or the like to obtain the surrounding condition. May be sought.

The voice recognition result extraction unit 24 extracts conversion candidates suitable for the conversation situation determined by the situation determination unit 23 from the plurality of conversion candidates obtained by the voice recognition unit 22. The extraction of conversion candidates will be described later.

The output unit 25 outputs conversion candidates (text) extracted by the voice recognition result extraction unit 24. The output also includes display on the display unit 16, notification to the control unit 11, transmission to the communication connection destination via the network 3, transmission as control information to various devices connected to the voice recognition device 1, and the like. It shall include.

Next, the voice recognition process executed by the voice recognition device 1 will be described with reference to FIG.
The control unit 11 reads out the program and data related to the voice recognition process shown in FIG. 3 from the storage unit 12, and executes the process based on the program and data.

First, the control unit 11 (voice acquisition unit 21) acquires the voice data of the conversation (step S101). The voice data may be input from the microphone 18 or may be input from a device communicatively connected to the voice recognition device 1 via the communication control unit 17 and the network 3. The control unit 11 (voice recognition unit 22) performs voice recognition on the acquired voice data (step S102). In step S102, the control unit 11 (voice recognition unit 22) performs a process of recognizing the voice of the conversation included in the voice data and converting it into text. The control unit 11 (speech recognition unit 22) obtains one or a plurality of conversion candidates as a result of the voice recognition process. When there are a plurality of conversion candidates, the conversation status is determined by the processes of steps S103 to S104.

The control unit 11 (conversation content determination unit 231 of the situation determination unit 23) determines the conversation content (genre, etc.) as the conversation status (step S103). The control unit 11 (conversation content determination unit 231) obtains the conversation content by analyzing the meaning of the word included in the voice recognition result (conversion candidate) in step S102. The content of the conversation requested here is the genre or purpose of the conversation. The conversation content determination unit 231 determines the content (genre) of the conversation by referring to the genre data 5 stored in advance in the storage unit 12, for example.

As shown in FIG. 4, the genre data 5 is data in which one or a plurality of conversion candidate words for word reading (speech recognition result) are associated with the genre of the word. For example, the conversion candidates for the voice recognition result "Taiken" include "Daiken" and "Experience", the genre of the conversion candidate "Daiken" is "RPG (game)", and the genre of the conversion candidate "Experience" is. It is "sightseeing". In this way, one or a plurality of conversion candidates and genres corresponding to each conversion candidate are stored for one voice recognition result. Multiple genres may be associated with each conversion candidate. The conversation content determination unit 231 obtains one or a plurality of conversation content candidates (genre candidates) for the voice recognition result by referring to the genre data 5. For example, if the voice recognition result is "Do you have a recommended Taiken?", The word "Taken" is included, so "RPG (game)" and "sightseeing" are required as genre candidates.

Next, the control unit 11 (surrounding situation determination unit 232 of the situation determining unit 23) determines the surrounding situation (step S104). The control unit 11 (surrounding condition determination unit 232) analyzes the characteristics of the surrounding sound when the voice data of the conversation is input, and obtains the surrounding condition such as a place based on the characteristics of the sound. For example, the sounds recorded in the "restaurant" include features such as the sound of tableware and the sound of customer service. The surrounding situation determination unit 232 stores in advance the data associated with the sound feature and the situation in the storage unit 12 as sound feature data (not shown), and determines the surrounding situation of the conversation based on the sound feature data. You may do so. The control unit 11 (surrounding condition determination unit 232) seeks one or a plurality of candidates for surrounding conditions. The surrounding situation is not limited to the place, and may be the purpose of conversation such as "customer service", "class", "chat" and the like. The control unit 11 obtains, for example, a "tourist information center", a "customer service", etc. as a result of determining the surrounding situation by analyzing the characteristics of the sound.

The control unit 11 (surrounding condition determination unit 232) may determine the surrounding condition based on not only the characteristics of the surrounding sound but also the image and the position information. Specifically, the image analysis unit 28 is provided with an image analysis unit 28 that acquires and analyzes a video (image) of a conversation taken by an image acquisition unit 27 (camera or the like), and the image analysis unit 28 inputs conversation audio data. You may ask for the surrounding situation, such as the place and purpose of the conversation when the conversation was held. Further, the surrounding situation may be obtained by acquiring the position information such as GPS and obtaining the place of conversation (store or facility) or the like based on the position information and the map data stored in advance.

The control unit 11 (speech recognition result extraction unit 24) has the conversation status (conversation content (genre) and surrounding situation) obtained in steps S103 and S104 among the plurality of conversion candidates of the voice recognition result obtained in step S102. Conversion candidates suitable for the above are extracted (step S105). For example, if the voice recognition result is "Do you have a recommended taiken?", The word "taiken" included in this sentence has conversion candidates of "big sword" and "experience". When the content of the conversation is determined to be "RPG (game)", "sightseeing", ... In step S103, and the surrounding situation is determined to be "sightseeing", "customer service", ... In step S104, the control unit 11 talks. Match the genre of the content with the surrounding situation and extract words of the genre with high probability. In the above example, since "sightseeing" is judged to be plausible as the conversation situation, the conversion candidate "experience" is selected, and "Do you have a recommended experience?" Is obtained as the voice recognition result of the input voice.

The control unit 11 stores the situation determined in steps S103 and S104 (“sightseeing” in the above example) as situation history data in the storage unit 12 in association with time information (input time of voice data, etc.) (step S106). ).

The control unit 11 (output unit 25) outputs the voice recognition result extracted in step S105. The output also includes display on the display unit 16, notification to the control unit 11, transmission to the communication connection destination via the network 3, transmission of control signals to various devices connected to the voice recognition device 1, and the like. It shall include.

When the voice recognition result is output, the voice recognition process for the input voice data is terminated.

In the above-mentioned voice recognition process, the conversation content may be determined based on the history saved in step S106. That is, in step S103, when the history of the conversation status up to the previous sentence is saved, the control unit 11 may narrow down the conversation content (genre) from the conversation status up to the previous sentence.

For example, as shown in FIG. 5, the voices of conversations such as "Welcome" and "Are you looking for something" are input at 17:50, and the voices at 17:50 ("Welcome", "Something"). As for the situation of "Are you looking for?"), Conversation genres such as "store", "family restaurant", and "drug store" are required. It is assumed that each candidate of the conversation genre is given a likelihood. For example, the likelihood of the conversation genre "store" for the word "welcome" is "1.0", the likelihood of "family restaurant" is "0.9", and the likelihood of "drugstore" is "0". .9 ", ... and so on. It is assumed that the likelihood is given in advance to the genre data 5 in which words and genres are associated with each other, for example. In addition, as a situation determination result by sound analysis at 17:50, "tourist information center" and "reception" are obtained together with the likelihood. For example, the likelihood of "tourist information center" is calculated as "0.8", and the likelihood of "reception" is calculated as "0.5". The likelihood of the situation determination result by this sound analysis may be given from the degree of coincidence between the input sound and the sound feature data stored in advance.

Next, at 17:56, the voice of the conversation "Where is the XX leisure facility?" And "It is 5 minutes from here" is input. The control unit 11 determines that the genre of conversation at 17:56 is "amusement park", "sightseeing", or the like. The likelihood of each is "0.9" for "amusement park" and "0.8" for "sightseeing". In addition, as the situation judgment result by sound analysis at 17:56, "tourist information center", "storefront", and "customer service" are obtained. It is assumed that the likelihood of "tourist information center" is "0.8", the likelihood of "storefront" is "0.8", and the likelihood of "customer service" is "0.7".

After that, at 18:00, it is assumed that the voice "Do you have a recommended taken?" To be processed is input. The control unit 11 determines that the genre of conversation at 18:00 is "RPG (game)", "sightseeing", or the like. It is assumed that "RPG (game)" obtains "0.5" and "sightseeing" obtains "0.2" as their respective likelihoods. In addition, "tourist information center" and "customer service" are obtained as the situation judgment result by sound analysis at 18:00. It is assumed that the likelihood of "tourist information center" is "0.8" and the likelihood of "customer service" is "0.7".

The likelihood based on the conversation content of each conversion candidate ("big sword", "experience") of "taiken" is "0.5" for RPG (game) and "0.2" for sightseeing, but the sound In the situation judgment by, there is no candidate for RPG (game) (likelihood "0"), and tourism (tourist information center) has a likelihood "0.8". Taken together,
Likelihood of "RPG (game)" = conversation content "0.5" + sound judgment "0" = 0.5
Likelihood of "sightseeing" = conversation content "0.2" + sound judgment "0.8" = 1.0
Will be.

Therefore, as a situation, the likelihood of "sightseeing" is higher than that of "RPG (game)" and it is appropriate. Therefore, the control unit 11 determines that "experience" is suitable as the conversion result (speech recognition result) of "taiken". In this way, if the history of voice recognition and sound judgment is traced back and used for situation judgment, it is possible to obtain a voice recognition result in consideration of the context (context, flow of conversation). For example, if you look at the history of recent "RPG (game)" and "sightseeing" contexts, many "sightseeing" will appear, so this value (= weight) is multiplied by the likelihood of voice recognition and sound judgment and added. Can also be compared and judged.
Likelihood of "RPG (game)" = (conversation content "0.5" + sound judgment "0") x game context weight "0" = 0
Likelihood of "sightseeing" = (conversation content "0.2" + sound judgment "0.8") x tourism context weight "0.7" = 0.7

As described above, the voice recognition device 1 of the present embodiment recognizes the voice data of the conversation, obtains a plurality of conversion candidates as the voice recognition result, and considers not only the voice recognition result but also the ambient sound and the like. By judging the conversation situation, a candidate suitable for the conversation situation is extracted from a plurality of conversion candidates. As a result, it becomes possible to obtain a voice recognition result suitable for the conversation situation, and it is possible to improve the voice recognition accuracy.

When the voice recognition device 1 of the present invention is applied to a communication device such as a smartphone 2 or a tablet, or a PC (Personal Computer) 7 connected to a network 3 such as the Internet, the voice recognition system 100 shown in FIG. As described above, each functional unit (speech acquisition unit 21, voice recognition unit 22, situation determination unit 23, voice recognition result extraction unit 24, output unit 25) of the voice recognition device 1 according to the present invention can access the server accessible from the smartphone 2 or the like. , Sound analysis unit 26, etc.) may be provided. That is, when the smartphone 2, PC 7, and the like transmit the voice data of the conversation input from the microphone 18 to the voice recognition device 1 (server) via the network 3, the voice recognition device 1 executes the voice recognition process shown in FIG. Then, the voice recognition result may be returned to the voice input source smartphone 2 or the like.

Further, the voice recognition device 1 according to the present invention may be applied to the interactive robot 6. In this case, each function of the voice recognition device 1 according to the present invention (voice acquisition unit 21 (microphone), voice recognition unit 22, situation determination unit 23, voice recognition result extraction unit 24, output unit 25, sound analysis unit 26, image The robot 6 includes an acquisition unit 27 (camera), an image analysis unit 28, and a position information acquisition unit 29). Alternatively, the voice or image input from the voice acquisition unit 21 (microphone) and the image acquisition unit 27 (camera) of the interactive robot 6 is transmitted to the server (voice recognition device 1), and the server (voice recognition device 1) sends the voice or image. , The voice recognition process shown in FIG. 3 may be executed, and the voice recognition result may be returned to the interactive robot 6 that is the voice input source.

In addition, the voice recognition device 1 according to the present invention can also be applied to various information devices such as car navigation systems, home appliances, and the like.

Although preferred embodiments of the voice recognition device and the like according to the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above examples. It is clear that a person skilled in the art can come up with various modifications or modifications within the scope of the technical idea disclosed in the present application, and these also naturally belong to the technical scope of the present invention. Understood.

1 ……………… Voice recognition device 10 ……………… Computer 11 ……………… Control unit 12 ……………… Storage unit 13 ……………… Media input / output unit 14 ……………… Peripheral Equipment I / F section 15 ……………… Input section 16 ……………… Display section 17 ……………… Communication control section 18 ……………… Microphone 19 ……………… Bus 21 ……………… Voice acquisition unit 22 ……………… Voice recognition unit 23 ……………… Situation judgment unit 231 ………… Conversation content judgment unit 232 ………… Surrounding situation judgment unit 24 ……………… Voice recognition result extraction unit 25 ……………… Output unit 26 ……………… Sound analysis unit 27 ……………… Image acquisition unit 28 ……………… Image analysis unit 29 ……………… Position information acquisition unit 3 ………… …… Network 5 ……………… Genre data 6 ……………… Interactive robot 7 ……………… PC
100 ………… Voice recognition system

Claims (8)

Voice acquisition means to acquire voice data and
A voice recognition means that recognizes the acquired voice data and seeks multiple conversion candidates as the voice recognition result.
Situation determination means for determining the situation at the time of voice data acquisition,
An extraction means for extracting a conversion candidate suitable for the situation determined by the situation determination means from a plurality of conversion candidates obtained by the voice recognition means, and an extraction means.
Equipped with a memory means to memorize the words associated with the situation
The situation determining means, the word extracted from the voice data, the voice recognition device comprising a call determines the status based on the word. The voice recognition device according to claim 1 , wherein the situation determination means further determines the situation based on the acquired ambient sound. The voice recognition device according to claim 1 or 2 , wherein the situation determination means further determines the situation based on the acquired image. The situation determining means further speech recognition apparatus according to any one of claims 1 to 3, characterized in that determining the status based on the position information obtained. Voice acquisition means to acquire voice data and
A voice recognition means that recognizes the acquired voice data and seeks multiple conversion candidates as the voice recognition result.
Situation determination means for determining the situation at the time of voice data acquisition,
An extraction means for extracting a conversion candidate suitable for the situation determined by the situation determination means from a plurality of conversion candidates obtained by the voice recognition means, and an extraction means.
A storage means to store the status of the conversation history determined by said condition determining means,
Bei to give a,
The extraction means is characterized in that, based on the history of the conversation situation stored in the storage means, a conversion candidate suitable for the conversation situation is extracted from a plurality of conversion candidates obtained by the voice recognition means. Ruoto voice recognition device. The voice recognition device according to any one of claims 1 to 5 , wherein the extraction means prioritizes extraction candidates based on a set likelihood. The computer
Steps to get audio data and
The step of recognizing the acquired voice data and obtaining multiple conversion candidates as the voice recognition result,
Steps to determine the situation at the time of voice data acquisition,
A step of extracting a conversion candidate suitable for the situation from the plurality of conversion candidates, and
Memorize the words associated with the situation Memorize steps and
Including
Determining the status, the words extracted from the speech data, a speech recognition method comprising the this determines the status based on the word. Computer,
Voice acquisition means for acquiring voice data,
A voice recognition means that recognizes the acquired voice data and seeks multiple conversion candidates as the voice recognition result.
Situation determination means for determining the situation at the time of voice data acquisition,
An extraction means for extracting a conversion candidate suitable for a situation determined by the situation determination means from a plurality of conversion candidates obtained by the voice recognition means.
A mnemonic that memorizes words associated with a situation,
A program to function as,
The situation determination means is a program characterized in that the word is extracted from voice data and the situation is determined based on the word.

Images