JP7035979B2 - Speech recognition device - Google Patents

Description

The present invention relates to a voice recognition device.

Patent Document 1 discloses an automatic response recording device that reproduces the opening guidance and the utterance promotion guidance to the user and determines the end of the call from the user's utterance status after the reproduction of the utterance promotion guidance.

Japanese Unexamined Patent Publication No. 5-110690

In the apparatus described in Patent Document 1, the end-of-call determination is performed when the silent period in which no voice is input continues for a predetermined period (for example, 1 second), and the voice recognition process is started after the end-of-call determination. Therefore, the silent period until the end of the call is determined becomes the waiting time of the user as it is. Therefore, the device described in Patent Document 1 has a problem that the waiting time of the user cannot be shortened to the end-of-call determination period (= silence period) or less, no matter how fast the voice recognition process is.

The present invention has been made in view of the above, and an object of the present invention is to provide a speech recognition device capable of shortening a user's waiting time in a speech recognition process and improving the response of the speech recognition process.

In order to solve the above-mentioned problems and achieve the object, the voice recognition device according to the present invention is a voice recognition device provided with a control unit for processing the input voice, and the control unit is such that the voice is used. The end-of-call determination is performed when the silent period, which is a non-input period, elapses from the second threshold period, and the voice recognition process is started when the silent period elapses from the first threshold period shorter than the second threshold period. ..

As a result, the voice recognition device starts the voice recognition process without waiting for the end-of-call determination because the first threshold period for starting the voice recognition process is shorter than the second threshold period for determining the end of the call.

Further, in the voice recognition device according to the present invention, the control unit interrupts the voice recognition process when the silence period ends after the lapse of the first threshold period and before the lapse of the second threshold period. You may.

As a result, even if the voice recognition process is started once, the voice recognition device interrupts the voice recognition process if the user speaks before the elapse of the second threshold period.

According to the voice recognition device according to the present invention, since the voice recognition process is started without waiting for the end-of-call determination, the waiting time of the user in the voice recognition process can be shortened and the response of the voice recognition process can be improved. Can be done.

FIG. 1 is a functional block diagram showing a configuration of a voice recognition device according to an embodiment of the present invention. FIG. 2 is a diagram showing the timing of starting the voice recognition process in the conventional voice recognition device. FIG. 3 is a diagram showing the start timing of the voice recognition process in the voice recognition device according to the embodiment of the present invention. FIG. 4 is a time chart showing a flow of voice recognition processing when a user utters one sentence at a time without a gap in the voice recognition device according to the embodiment of the present invention. FIG. 5 is a time chart showing a flow of voice recognition processing when a user utters a plurality of sentences and words with a gap in the voice recognition device according to the embodiment of the present invention.

The voice recognition device according to the embodiment of the present invention will be described with reference to the drawings. The present invention is not limited to the following embodiments. In addition, the components in the following embodiments include those that can be easily replaced by those skilled in the art, or those that are substantially the same.

[Voice recognition device]
The voice recognition device 1 according to the present embodiment is mounted on a vehicle, for example, as an on-board unit, and includes a microphone 10, a control unit 20, and a voice buffer 30 as shown in FIG. The function of the voice recognition device 1 may be realized by a single device, or may be realized by a plurality of devices.

The microphone 10 collects the voice spoken by the user and outputs the voice signal to the voice acquisition unit 21 of the control unit 20. Specifically, the control unit (processor) 20 is composed of an arithmetic processing unit such as a CPU (Central Processing Unit), and processes voice (voice signal) input through the microphone 10. The control unit 20 includes a voice acquisition unit 21, a second end-of-call determination unit 22, a voice recognition unit 23, and a first end-of-call determination unit 24.

The voice acquisition unit 21 generates voice data by digitizing a time-series voice signal input from the microphone 10. Then, the voice acquisition unit 21 stores the generated voice data in the voice buffer 30. Further, the voice acquisition unit 21 outputs the voice data stored in the voice buffer 30 to the second end-of-call determination unit 22 and the voice recognition unit 23, if necessary.

The second end-of-call determination unit 22 makes a second end-of-call determination. Specifically, the second end-of-call determination unit 22 determines whether or not the silent period, which is a period during which no voice from the user is input, has elapsed a preset second threshold period. Then, the second end-of-call determination unit 22 makes a second-stage end-of-call determination (second end-of-call determination) when the silence period elapses from the second threshold period. The above-mentioned "second threshold period" is a period used as a threshold for determining whether or not the user has completely finished speaking. The second threshold period is longer than the above-mentioned first threshold period, for example, the period between the first end-of-call determination and the second end-end determination (for example, 1 second) is added to the first threshold period. It is set to the combined length (see FIG. 3).

The speech recognition unit 23 is a speech recognition engine that performs automatic speech recognition (ASR: Automatic Speech Recognition) processing (hereinafter referred to as "recognition processing"). The voice recognition unit 23 starts the recognition process when the silence period elapses from the first threshold value period. That is, the voice recognition unit 23 starts the recognition process when the first end-of-call determination is made by the first end-of-call determination unit 24.

Here, in the conventional voice recognition method, as shown in FIG. 2, the end-of-call determination is performed when the silence period elapses from a predetermined threshold period (for example, 1 second), and from the time when the end-of-call determination is performed. The recognition process was started. That is, in the conventional voice recognition method, since the end-of-call determination and the recognition process are sequentially processed (sequentially), there is a problem that the waiting time of the user becomes long.

On the other hand, in the voice recognition device 1 according to the present embodiment, as shown in FIG. 3, when the first threshold period in which the silence period is shorter than the conventional threshold period elapses, the first end-of-call determination is performed, and this first determination is made. The recognition process is started from the time when one end-of-call determination is made. That is, in the voice recognition device 1 according to the present embodiment, the recognition process is started earlier than the conventional voice recognition method, and the end-of-call determination and the recognition process are processed in parallel, thereby shortening the waiting time of the user. Aim for conversion.

The voice recognition unit 23 interrupts the recognition process when the silence period ends after the lapse of the first threshold period and before the lapse of the second threshold period, for example, when the silence period ends at the time point A in FIG. do. That is, in the voice recognition device 1 according to the present embodiment, even if the recognition process is started once, if the user speaks before the lapse of the second threshold period, the recognition process is interrupted.

The first end-of-speech determination unit 24 makes the first end-of-speech determination. Specifically, the first end-of-call determination unit 24 determines whether or not the silence period, which is the period during which no voice from the user is input, has elapsed the preset first threshold value period. Then, the first end-of-call determination unit 24 makes the end-of-call determination (first end-of-call determination) in the first stage when the silence period elapses from the first threshold value period. The above-mentioned "first threshold period" is used as a threshold value for determining whether or not the user has tentatively ended the call before determining whether or not the user has completely terminated the call (second end-of-call determination). It is a period to be. The first threshold period is shorter than the second threshold period described later, for example, the length obtained by subtracting the period (for example, 1 second) between the first end-of-call determination and the second end-end determination from the second threshold period. It is set to the threshold value (see FIG. 3 described later).

The audio buffer 30 is composed of, for example, an HDD (Hard Disk Drive), a ROM (Read Only Memory), a RAM (Random access memory), or the like. The voice buffer 30 temporarily stores the voice data generated by the voice acquisition unit 21.

[Voice recognition method]
Hereinafter, specific examples of the voice recognition method by the voice recognition device 1 according to the present embodiment will be described with reference to FIGS. 4 and 5. In the following description, the processing in the microphone 10, the voice acquisition unit 21, the second end call determination unit 22, and the voice buffer 30 will be described as the client-side processing, and the voice recognition unit 23 and the first end call determination unit 24 will be described. The process will be described as a process on the recognition engine side. Further, the "user operation" on the left side in FIGS. 4 and 5 indicates an action on the user side.

(First Example)
FIG. 4 shows the flow of the recognition process when the user utters one sentence (“call Mr. Sato”) at once to the voice recognition device 1 without a gap. First, when PTT (Push to Talk) is performed by the user (step S1), the voice recognition unit 23 on the recognition engine side is in an utterance waiting state (step S2).

Subsequently, when the user starts uttering "Call Mr. Sato" (step S3), the voice acquisition unit 21 on the client side starts accumulating voice data in the voice buffer 30 (step S4). At the same time, the voice recognition unit 23 on the recognition engine side detects the utterance (step S5).

Subsequently, when the user's utterance is completed (step S6) and the first threshold period elapses, the first end-of-call determination unit 24 on the recognition engine side performs end-of-call detection (first end-of-call determination) (step). S7). In response to this, the voice recognition unit 23 reads the voice data from the voice buffer 30 and starts the recognition process (step S8). Further, the first end-of-call determination unit 24 transmits the detection result to the second end-of-call determination unit 22 on the client side after the end-of-call determination (first end-of-call determination) (step S9). In response to this, the second end-of-call determination unit 22 waits for the complete end of the call (step S10).

Subsequently, when the second threshold period elapses from the end of the user's utterance, the second end-of-call determination unit 22 on the client side performs complete end-of-call detection (second end-of-call determination) (step S11). After that, when the recognition process in the voice recognition unit 23 is completed, the voice recognition unit 23 transmits the recognition result to the client side (step S12).

(Second Example)
FIG. 5 shows a time showing a flow of voice recognition processing when a user speaks a plurality of sentences (“search nearby”) and words (“convenience store”) with respect to the voice recognition device 1 with a gap. It is a chart. First, when PTT (Push to Talk) is performed by the user (step S21), the voice recognition unit 23 on the recognition engine side is in an utterance waiting state (step S22).

Subsequently, when the user starts the utterance "search nearby" (step S23), the voice acquisition unit 21 on the client side starts accumulating voice data in the voice buffer 30 (step S24). At the same time, the voice recognition unit 23 on the recognition engine side detects the utterance (step S25).

Subsequently, when the user's utterance is interrupted (step S26) and the first threshold period elapses, the first end-of-call determination unit 24 on the recognition engine side performs end-of-call detection (first end-of-call determination) (step S27). ). In response to this, the voice recognition unit 23 reads the voice data from the voice buffer 30 and starts the recognition process (step S28). Further, the first end-of-call determination unit 24 transmits the detection result to the second end-of-call determination unit 22 on the client side after the end-of-call detection (first end-of-call determination) (step S29). In response to this, the second end-of-call determination unit 22 waits for the complete end of the call (step S30).

Subsequently, when the user's utterance is resumed and the user starts the utterance of "convenience store" (step S31), the voice acquisition unit 21 on the client side re-detects the utterance (step S32). Then, the voice acquisition unit 21 transmits an instruction to interrupt the recognition process to the voice recognition unit 23 on the recognition engine side (step S33). In response to this, the voice recognition unit 23 interrupts the voice recognition. Further, the voice acquisition unit 21 transmits the voice data stored in the voice buffer 30 to the voice recognition unit 23 (step S34).

Subsequently, when the user's utterance is completed (step S35) and the first threshold period elapses, the first end-of-call determination unit 24 on the recognition engine side performs end-of-call detection (first end-of-call determination) (step). S36). In response to this, the voice recognition unit 23 reads the voice data from the voice buffer 30 and starts the recognition process (step S37). Further, the first end-of-call determination unit 24 transmits the detection result to the second end-of-call determination unit 22 on the client side after the end-of-call detection (first end-of-call determination) (step S38). In response to this, the second end-of-call determination unit 22 waits for the complete end of the call (step S39).

Subsequently, when the second threshold period elapses from the end of the user's utterance, the second end-of-call determination unit 22 on the client side performs complete end-of-call detection (second end-of-call determination) (step S40). After that, when the recognition process in the voice recognition unit 23 is completed, the voice recognition unit 23 transmits the recognition result to the client side (step S41).

According to the voice recognition device 1 as described above, the end-of-call determination in the recognition process is divided into two stages, and the end-of-call determination in a short time is performed before the end-of-call determination of a normal length (second end-end determination). By carrying out the talk determination (first end-of-speech determination), the recognition process can be started at an early stage.

That is, according to the voice recognition device 1, since the first threshold period for starting the recognition process is shorter than the second threshold period for determining the end of call, the recognition process is started without waiting for the end of call determination. Therefore, according to the voice recognition device 1, since the voice recognition process is started without waiting for the end call determination, the waiting time of the user in the voice recognition process can be shortened, and the response of the voice recognition process can be improved. can.

[Voice recognition program]
In the voice recognition program according to the present embodiment, the computer is made to function as each part (each means) of the control unit 20 described above. The voice dialogue program may be stored and distributed in a computer-readable recording medium such as a hard disk, a flexible disk, or a CD-ROM, or may be distributed via a network.

The voice recognition device according to the present invention has been specifically described above in terms of the mode for carrying out the invention, but the purpose of the present invention is not limited to these descriptions, but is based on the description of the scope of claims. Must be widely interpreted. Needless to say, various changes, modifications, etc. based on these descriptions are also included in the gist of the present invention.

1 Voice recognition device 10 Microphone 20 Control unit 21 Voice acquisition unit 22 Second end-of-speech judgment unit 23 Voice recognition unit 24 First end-of-speech judgment unit 30 Voice buffer

Claims (2)

A voice recognition device equipped with a control unit that processes input voice.
The control unit
When the silence period, which is the period during which no voice is input, elapses from the second threshold period, the end-of-call determination is performed.
The voice recognition process is started when the first threshold period shorter than the second threshold period elapses.
Speech recognition device. The voice recognition device according to claim 1, wherein the control unit interrupts the voice recognition process when the silence period ends after the lapse of the first threshold period and before the lapse of the second threshold period.

Images