JP2017181667A - Voice recognition apparatus and voice recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice recognition apparatus with its recognition accuracy and speed well balanced.SOLUTION: A voice recognition apparatus, for use in recognizing a voice obtained from a user, includes: voice obtainment means for obtaining a voice from the user; recognition means for recognizing the voice and obtaining a first recognition result and a degree of certainty corresponding to the first recognition result; information obtainment means for obtaining a second recognition result that is a result of recognizing the voice by a second voice recognition apparatus being different from the apparatus itself; and determination means for adopting the first recognition result in a case where the degree of certainty when the voice recognized by the recognition means is equal to or higher than a given value, or making an inquiry to the user in a case where the degree of certainty is less than the given value and, based on the response to the inquiry, determining which of the first recognition result and second recognition result to be adopted.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a speech recognition apparatus that recognizes input speech.

  A speech recognition technology is widely used in which a speech uttered by a user is recognized and a computer performs processing using the recognition result. By using the voice recognition technology, it is possible to operate the computer without contact, and the convenience of the computer mounted on a moving body such as an automobile is greatly improved.

The recognition accuracy when performing speech recognition varies depending on the scale of the dictionary used for recognition. For example, there may be a large difference in recognition accuracy between a workstation specialized for speech recognition and a personal computer not specialized for speech recognition.
Thus, when it is desired to use speech recognition in a small-scale computer, a technique of transferring speech data to a large-scale computer through a communication line and acquiring a recognition result is used.

However, when speech recognition is performed via a network, there is a problem that the overall response becomes slow. That is, there is a problem in terms of both recognition accuracy and response.
In order to cope with this, a technique for dynamically switching whether voice recognition is performed locally or on a server has been proposed. For example, Patent Literature 1 describes a system that calculates the reliability of a locally recognized speech recognition result and adopts the speech recognition result by a server when the reliability is low.

JP2013-232001A

  In the system described in Patent Document 1, the reliability of the result of speech recognition performed locally is compared with a threshold value, and the result of performing speech recognition locally is used based on the comparison result, or speech recognition is performed by a server. Have decided. However, if the determination is made depending only on the threshold value stored locally, a case where an appropriate determination cannot be made occurs.

For example, if the threshold is too high with respect to the scale of the dictionary, voice recognition will always be performed by the server, and the overall response will be reduced. If the threshold is too low, the response is fast, but a high quality recognition result cannot be obtained.
Also, if the reliability is below the threshold, the result of speech recognition performed by the server unconditionally is adopted even if the result of speech recognition performed locally is correct.
That is, there is a problem in terms of achieving both recognition accuracy and speed.

  The present invention has been made in consideration of the above-described problems, and an object thereof is to provide a speech recognition apparatus that achieves both recognition accuracy and speed.

The speech recognition apparatus according to the present invention is
A speech recognition apparatus for recognizing speech acquired from a user, wherein speech acquisition means for acquiring speech from the user, recognition of the speech, a first recognition result, and a probability corresponding to the first recognition result Recognizing means, information obtaining means for obtaining a second recognition result obtained by causing a second voice recognition device different from the device to recognize the voice, and the recognition means recognizes the voice. When the accuracy at the time is greater than or equal to a predetermined value, the first recognition result is adopted, and when the accuracy is lower than the predetermined value, an inquiry is made to the user, and based on an answer to the inquiry, the first Determining means for deciding which one of the recognition results or the second recognition result is to be adopted.

  The first recognition result is a result of performing speech recognition locally, and the second recognition result is a result of performing speech recognition using a speech recognition device different from the own device (that is, located outside). is there.

  Speech recognition is preferably performed locally from the viewpoint of response time if a certain recognition accuracy can be obtained, and from the viewpoint of accuracy if an external device is used if a certain recognition accuracy cannot be obtained. preferable. That is, it is possible to determine whether to adopt the first recognition result or the second recognition result using the accuracy of the local speech recognition result. However, if you decide which one to use based only on the accuracy, for example, even if the first recognition result is correct as a result, extra communication may occur when trying to adopt the second recognition result. The problem that response time becomes long occurs.

Therefore, the speech recognition apparatus according to the present invention adopts the first recognition result unconditionally when the accuracy of the first recognition result is equal to or higher than the predetermined value, and the accuracy is lower than the predetermined value. And determines whether to adopt the first recognition result or the second recognition result based on the contents of the response to the inquiry.
The contents of the inquiry to the user are typically related to the correctness of the first recognition result, but are not limited thereto. According to such a configuration, the user is inquired only when there is doubt about the local recognition result, and as a result, only when it is recognized that there is a clear error, voice recognition is performed using an external source. It is possible to reduce the probability of adopting an incorrect recognition result while taking advantage of the responsiveness of performing voice recognition locally.

  Further, the determination means inquires of the user whether the first recognition result is correct when the accuracy is lower than a predetermined value, and adopts the first recognition result when the answer is affirmative. It may be characterized by.

  When the accuracy falls below a predetermined value, the user is inquired and the first recognition result is adopted when there is an affirmative answer, that is, when the first recognition result is correct. By doing so, although the accuracy is relatively low, the response speed can be secured in the case where the result of the speech recognition performed locally is correct as a result.

  Further, the determination means may adopt the second recognition result when the answer is negative and the first recognition result and the second recognition result are different.

  This is because if the user's answer is negative and the first recognition result is different from the second recognition result, the second recognition result is likely to be correct.

  Further, the determination means inquires of the user whether the second recognition result is correct when the answer is negative and the first recognition result is different from the second recognition result. The second recognition result may be adopted when the result is positive.

  According to this configuration, it is possible to determine whether to adopt the second recognition result or re-perform the recognition itself.

  Further, the information acquisition means transmits the voice acquired by the voice acquisition means to the second voice recognition device, and acquires the second recognition result when the answer is negative. It is good.

The second recognition result is not necessarily acquired at the same time as the first recognition result. For example, the response can be speeded up by not obtaining the second recognition result until a negative answer to the inquiry is obtained.
Note that the timing for transmitting data to the second speech recognition apparatus may be the timing for acquiring the first recognition result, or may be after determining that the first recognition result is not adopted. Moreover, you may perform the inquiry with respect to a 2nd speech recognition apparatus in parallel.

  The second voice recognition device may be a device that communicates with the determination unit via a wireless network, and may be a device having higher voice recognition accuracy than the own device.

  In such a configuration, a trade-off problem between recognition accuracy and response occurs, and the present invention can be suitably applied.

  The speech recognition apparatus according to the present invention may further include information providing means for providing information to the user based on a speech recognition result adopted by the determining means.

  The information providing means is means for providing information to the user based on the result of recognizing the voice. The provided information includes, for example, route information to the destination, information related to the route, web search results, search results for databases, search results for other servers, and the like, but is not limited thereto. Further, the computer may perform processing according to the voice recognition result, and the result of the processing may be provided as information.

  Note that the present invention can be specified as a speech recognition apparatus including at least a part of the above means. It can also be specified as a speech recognition method performed by the speech recognition apparatus. The above processes and means can be freely combined and implemented as long as no technical contradiction occurs.

  ADVANTAGE OF THE INVENTION According to this invention, the speech recognition apparatus which made recognition accuracy and speed compatible can be provided.

It is a system configuration figure of an information service system concerning an embodiment. It is a flowchart figure of the process which the information provision apparatus which concerns on embodiment performs. It is a flowchart figure of the process which the information provision apparatus which concerns on embodiment performs.

Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.
The information providing system according to the present embodiment acquires a voice command from an occupant (for example, a driver) boarding a vehicle, performs voice recognition, collects corresponding information based on the recognition result, and provides it to the occupant System.

<System configuration>
FIG. 1 is a system configuration diagram of an information providing system according to the present embodiment.
The information providing system according to this embodiment includes an information providing apparatus 10 and an information providing server 20.

First, the information providing apparatus 10 will be described. The information providing apparatus 10 is an apparatus having a function of acquiring voice from a vehicle occupant and performing voice recognition, a function of acquiring information based on a result of the voice recognition, and a function of providing the acquired information to a vehicle occupant. . For example, the information providing apparatus 10 may be a car navigation apparatus mounted on a vehicle or a general-purpose computer. Moreover, another vehicle-mounted terminal may be sufficient.
Further, the information providing apparatus 10 has a function of performing voice recognition processing using an external voice recognition apparatus (voice recognition server 20) instead of the own apparatus. Details of the voice recognition server 20 will be described later.

The information providing apparatus 10 includes a voice acquisition unit 11, a voice recognition unit 12, an information acquisition unit 13, an input / output unit 14, and a communication unit 15.
The sound acquisition unit 11 is means for acquiring sound from a vehicle occupant. Specifically, sound is converted into an electrical signal (hereinafter referred to as sound data) using a microphone (not shown). The acquired voice data is transmitted to the voice recognition unit 12.

  The voice recognition unit 12 is a means for performing voice recognition on the acquired voice data and converting it into text. Speech recognition can be performed by known techniques. For example, the speech recognition unit 12 stores an acoustic model and a recognition dictionary, extracts features by comparing the acquired speech data with the acoustic model, and matches the extracted features with the recognition dictionary to generate speech. Recognize. The speech recognition unit 12 is configured to be able to acquire the likelihood (that is, the accuracy) of the recognition result together with the recognition result (text).

The information acquisition unit 13 is means for acquiring information to be provided to the occupant based on the voice recognition result.
The information to be provided may be, for example, information representing a route connecting the departure point and the destination, information related to the route, or information related to the movement of the vehicle. In addition, information that is not directly related to vehicle travel, such as web search results, may be used. For example, when providing route information, a route search may be performed in response to a voice command “route to XX (destination)” and the result may be presented. Further, when providing information related to a route, for example, the required time may be estimated in response to a voice command “required time to XX (destination)” and the result may be presented. In addition to this, as long as information can be provided by natural language processing, the contents of the voice command and the information to be provided may be anything.

The input / output unit 14 is means for accepting an input operation performed by the user and presenting information to the user. In this embodiment, it consists of one touch panel display. That is, it comprises a liquid crystal display and its control means, a touch panel and its control means.
The communication unit 15 is means for communicating with the voice recognition server 20 by accessing the network via a communication line (for example, a mobile phone network).

The voice recognition server 20 is a server device specialized for voice recognition, and includes a communication unit 21 and a voice recognition unit 22.
Since the function which the communication part 21 has is the same as that of the communication part 15 mentioned above, detailed description is abbreviate | omitted. The speech recognition unit 22 has the same function as the speech recognition unit 12 described above, but differs in that the scale of the recognition dictionary it has is larger than that of the speech recognition unit 12. That is, the voice recognition unit 22 is configured to perform voice recognition with higher accuracy than the voice recognition unit 12.

  Each of the information providing apparatus 10 and the voice recognition server 20 can be configured as an information processing apparatus having a CPU, a main storage device, and an auxiliary storage device. Each unit shown in FIG. 1 functions by loading a program stored in the auxiliary storage device into the main storage device and executing it by the CPU. Note that all or part of the illustrated functions may be executed using a circuit designed exclusively.

<Process flowchart>
Next, the content of specific processing performed by the information providing apparatus 10 will be described. FIG. 2 is a flowchart illustrating processing executed by the information providing apparatus 10.
First, in step S11, the voice acquisition unit 11 acquires voice from a vehicle occupant through a microphone (not shown). The acquired voice is converted into voice data and transmitted to the voice recognition unit 12 and the communication unit 15.
Next, in step S12, the voice recognition unit 12 performs voice recognition on the acquired voice data and converts the voice into text. The text of the conversion result is transmitted to the information acquisition unit 13 together with the likelihood.

  Next, in step S <b> 13, the communication unit 15 starts transmitting the acquired voice data to the voice recognition server 20. Further, the information providing apparatus 10 performs the process of step S14 in parallel with the transmission of the audio data. The transmitted voice data is converted into text by the voice recognition unit 22, and transmitted to the information acquisition unit 13 via the communication unit 21 and the communication unit 15 as soon as the conversion is completed.

  In step S14, the information acquisition unit 13 determines whether or not local speech recognition has succeeded (that is, whether or not a speech recognition result has been obtained). As a result, when a voice recognition result is obtained, the process proceeds to step S15.

  In step S15, the information acquisition unit 13 determines whether or not the likelihood (hereinafter referred to as local likelihood) when the speech recognition unit 12 performs speech recognition is equal to or greater than a predetermined threshold. As a result, if the local likelihood exceeds the threshold value, the local speech recognition result is adopted and information to be provided is generated (step S16).

If the local likelihood is below the threshold value, the result of speech recognition performed by the speech recognition unit 12 is presented to the user through the input / output unit 14 to determine whether it is correct (step S17). The inquiry may be made through a screen or may be made by voice. The confirmation performed here may present the recognition result as it is, or may interpret the context based on the recognition result and present the result.
For example, the content of the inquiry may be “Are you sure that the command you entered is“ Tell me about restaurants near OO station? ”Or“ Search for restaurants around OO station ” "Are you sure?"

  As a result, if the answer is affirmative, that is, if an answer indicating that the recognition result is correct is obtained, the local speech recognition result is adopted and information to be provided is generated (step S16).

  On the other hand, when a negative answer is obtained in step S17, that is, when an answer indicating that the recognition result is incorrect is obtained, the information providing apparatus 10 displays the result of the voice recognition server 20 performing voice recognition. Acquire and try to hire.

This process will be described with reference to FIG. The process illustrated in FIG. 3A is a process performed by the information acquisition unit 13.
First, in step S21, it is determined whether there is a recognition result acquired from the voice recognition server 20. Here, when the recognition result is not acquired, it transfers to step S25 and performs reacquisition of an audio | voice. The case where the recognition result cannot be acquired includes a case where a response indicating that the recognition could not be performed is received, or a case where a response is not obtained even after the timeout time has elapsed.

  If there is a recognition result transmitted from the voice recognition server 20, it is determined whether or not the recognition result is the same as the local recognition result (step S22). As a result, if they are the same, both the local and server recognition results are incorrect, so the process proceeds to step S25 to reacquire voice.

If the local recognition result and the server recognition result are different from each other, the process proceeds to step S23, where the result of the voice recognition performed by the server is presented to the user for confirmation. The confirmation method may be the same as that performed in step S17.
As a result, when the answer is affirmative (that is, when an answer indicating that the recognition result is correct is obtained), the process proceeds to step S24, and the voice recognition result in the server is adopted. When a negative answer is obtained (that is, when an answer indicating that the recognition result is incorrect) is obtained, the process proceeds to step S25, and voice is reacquired.

On the other hand, if the local speech recognition has failed in step S14, the information providing apparatus 10 acquires the result of the speech recognition server 20 performing the speech recognition and tries to adopt it.
This process will be described with reference to FIG. The process illustrated in FIG. 3B is a process performed by the information acquisition unit 13.
First, in step S21, it is determined whether there is a recognition result acquired from the voice recognition server 20. Here, when the recognition result is not acquired, it transfers to step S25 and performs reacquisition of an audio | voice. On the other hand, when there is a recognition result transmitted from the voice recognition server 20, the process proceeds to step S24, and the voice recognition result in the server is adopted.
That is, the process of FIG. 3B is different from the process of FIG. 3A in that the process of presenting the recognition result in the server to the user is omitted.

As described above, the information providing apparatus according to the present embodiment allows the user to confirm the recognition result when the likelihood of the local speech recognition result is below the threshold, and the answer is negative. Only, the process which employ | adopts the speech recognition result in a server is performed.
According to such a configuration, when there is doubt about the local recognition result, rather than adopting the recognition result in the server unconditionally, in order to determine which to adopt after inquiring the user, Recognition accuracy can be ensured while improving responsiveness.

(Modification)
The above embodiment is merely an example, and the present invention can be implemented with appropriate modifications within a range not departing from the gist thereof.
For example, the threshold value in step S15 may not be a fixed value. For example, it may be a dynamic value set based on the learning result.
When two or more results are obtained as a result of voice recognition, inquiries may be made by presenting the contents in order in the process of step S17 or S23.

  Moreover, when the likelihood of speech recognition (hereinafter referred to as server likelihood) can be acquired from the speech recognition server, whether to execute Step S23 or omit it may be determined based on the server likelihood. For example, when the server likelihood exceeds a predetermined threshold, step S23 may be omitted and step S24 may be executed.

In the description of the embodiment, transmission of voice to the voice recognition server is started in step S13. However, transmission of voice to the voice recognition server may be performed after step S21. By doing in this way, the amount of communication can be reduced.

DESCRIPTION OF SYMBOLS 10 ... Information provision apparatus 20 ... Voice recognition server 11 ... Voice acquisition part 12, 22 ... Voice recognition part 13 ... Information acquisition part 14 ... Input / output part 15, 21 ... Communication department

Claims (8)

A speech recognition device for recognizing speech acquired from a user,
Voice acquisition means for acquiring voice from the user;
Recognizing means for recognizing the voice and obtaining a first recognition result and an accuracy corresponding to the first recognition result;
Information acquisition means for acquiring a second recognition result that is a result of causing the second voice recognition device different from the device to recognize the voice;
When the accuracy when the recognition unit recognizes the voice is equal to or higher than a predetermined value, the first recognition result is adopted, and when the accuracy is lower than the predetermined value, the user is inquired and the inquiry is made. Determining means for deciding whether to adopt the first recognition result or the second recognition result based on an answer to
Voice recognition device. The determination means inquires to the user whether the first recognition result is correct when the accuracy is lower than a predetermined value, and adopts the first recognition result when the answer is affirmative.
The speech recognition apparatus according to claim 1. The determination means adopts the second recognition result when the answer is negative and the first recognition result is different from the second recognition result.
The speech recognition apparatus according to claim 2. The determination means inquires of the user whether the second recognition result is correct when the answer is negative and the first recognition result is different from the second recognition result. Adopt the second recognition result if
The speech recognition apparatus according to claim 2. The information acquisition means transmits the voice acquired by the voice acquisition means to the second voice recognition device, and acquires the second recognition result when the answer is negative.
The voice recognition device according to claim 2. The second speech recognition device is a device that communicates with the determining means via a wireless network, and is a device that has higher speech recognition accuracy than the own device,
The speech recognition apparatus according to claim 1. Further comprising information providing means for providing information to the user based on the speech recognition result adopted by the determining means;
The speech recognition apparatus according to claim 1. A speech recognition method performed by a speech recognition device that recognizes speech acquired from a user,
An audio acquisition step of acquiring audio from the user;
A recognition step of recognizing the voice and obtaining a first recognition result and an accuracy corresponding to the first recognition result;
When the accuracy when recognizing the voice in the recognition step is greater than or equal to a predetermined value, the first recognition result is adopted, and when the accuracy is lower than the predetermined value, the user is inquired and the inquiry is made. Based on the answer to the above, either the first recognition result or the second recognition result which is a result of causing the second voice recognition device different from the voice recognition device to recognize the voice is adopted. To decide,
Speech recognition method.

Images