JP2807241B2 - Voice recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial application field) The present invention relates to a voice recognition device capable of reliably collecting a voice of a speaker and performing voice recognition.

(Prior Art) Recently, voice recognition devices have been incorporated into various devices, voices such as commands uttered by a speaker have been fetched via a microphone, input voice has been recognized, and various information has been recognized in accordance with the recognition results. Various attempts have been made to perform processing. For example, attention has been paid to the use of a speech recognition device as a document information input unit in a word processor, a control command input unit in a plant or the like, and an information input unit in various ticket vending machines.

By the way, in this type of voice input device, it is important to reliably input only the voice intended for information input without being adversely affected by ambient noise and the like, and to recognize the voice in order to improve its recognition performance. Issues. Although the speech recognition process is generally started in response to a recognition start command from the control unit (host computer), it is very difficult to properly set the start timing of the speech recognition process in an actual system. Therefore, conventionally, a proximity sensor such as an ultrasonic sensor is generally provided near a microphone for inputting a voice uttered by a speaker, and the proximity sensor detects the proximity state between the speaker's face (mouth) and the microphone. For example, when it is detected that the speaker's face is sufficiently close to the microphone, a recognition start command is issued, and then a message instructing the start of voice input is output. When the proximity sensor detects that the speaker's face is separated from the microphone, a message indicating that the face (mouth) is close to the microphone and speaks is output to ensure a good voice input state. Has become.

However, in this way, the proximity sensor detects the proximity state between the microphone and the face of the speaker, controls the operation of the voice recognition process, outputs a message related to the voice input to the speaker, and outputs the timing of the voice input. Despite instructing the input attitude and the input attitude, some speakers start speaking before the voice input start message is output even if they are not sufficiently close to the microphone, or even if they are sufficiently close to the microphone. It often happens.
In such a case, the voice input is performed before the voice recognition start command is given and the voice recognition process is started, so that the voice recognition device outputs the voice without the leading portion. Will be entered. As a result,
An erroneous recognition or recognition rejection occurs, and it is necessary to prompt the speaker to re-input the voice, which increases the burden not only on the speaker but also on the voice recognition device.

In particular, if the start of the utterance is only slightly earlier than the timing of the generation request, since the speaker himself does not know that the utterance timing is earlier, the leading part is also lost when re-inputting the voice. There is a problem that it often happens.

In order to solve such a problem, it is considered that the sensitivity of the proximity sensor is increased, and a recognition start command is given at a timing when the speaker approaches the microphone. However, if such measures are taken, it becomes impossible to take measures for starting the utterance in a state where the speaker does not approach the microphone sufficiently, which causes an increase in erroneous recognition and recognition rejection. It is also considered that the above-described high-sensitivity proximity sensor is used to detect a speaker approaching, and then issue an utterance request after a certain period of time. However, there is a problem that it cannot be detected whether or not the distance has sufficiently approached.

(Problems to be Solved by the Invention) As described above, in the related art, it is detected that the speaker is sufficiently close to the microphone, and the speech from the speaker is fetched without causing the head portion to be lost. It is very difficult to appropriately set the timing of the start of the recognition process and the like for recognition.

The present invention has been made in view of such circumstances,
It is an object of the present invention to provide a speech recognition apparatus capable of reliably inputting and recognizing a speech uttered by a speaker without causing a loss of a leading portion thereof.

(Means for Solving the Problems) According to the present invention, when a voice uttered by a speaker is input by a microphone provided on a housing and the voice is recognized, the utterance is provided by a proximity sensor provided near the microphone. A voice recognition device configured to detect a proximity state between a person's face and the microphone and output a message related to voice input to a speaker according to the proximity state, and speak to the housing separately from the proximity sensor. An auxiliary proximity sensor for detecting that a speaker has approached, and when the proximity state of the speaker is detected by the auxiliary proximity sensor, a voice recognition process for voice input from the microphone is started. It is characterized by the following.

That is, prior to prompting the timing of voice input or the like according to the detection result by the proximity sensor, the voice recognition process is started according to the detection result by the auxiliary proximity sensor.

(Operation) According to the present invention, the proximity of the speaker is detected by the auxiliary proximity sensor before prompting the timing of voice input or the like according to the detection result of the proximity sensor, and the voice recognition process is started according to the detection result. Therefore, even if the voice start timing of the speaker is early, it is possible to reliably capture the voice without recognizing the head portion thereof and perform voice recognition.

As a result, it is possible to suppress erroneous recognition and recognition reject utterance very simply and effectively, thereby improving the recognition efficiency.

Hereinafter, a speech recognition device according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic diagram of a main part of the apparatus of the embodiment, and FIG. 2 is a diagram showing a control processing sequence thereof.

In FIG. 1, reference numeral 1 denotes a speech recognition unit configured by employing various recognition methods. The voice recognition unit 1 starts a voice recognition process in response to a recognition start command from the control unit 2, performs a recognition process on a voice input from the microphone 3, and outputs a recognition result.

The voice recognition unit 1 and the control unit 2 are generally installed in a housing 4 of a main body of various information processing apparatuses, and the microphone 3 is provided on a so-called operation panel surface of the housing 4. Can be

The operation panel provided with the microphone 3 has a display 5 for displaying various messages related to voice input to the speaker A, and a speaker A for the microphone 3.
A proximity sensor (first proximity sensor) 6 for detecting the proximity of the face (mouth) is provided at a position near the microphone 3. The control unit 2 basically uses the proximity sensor (first proximity sensor) 6 to determine the face (mouth) of the speaker A.
Of the microphone 3 is detected, and various messages relating to voice input are presented and output using the display 5 according to the detection result (proximity state). The message displayed on the display 5 is, for example, “Start voice input. "Please speak closer to the microphone and speak. ] Message. Of course, it is also possible to output these messages by voice using a voice synthesizer or the like.

The present apparatus is characterized in that an auxiliary proximity sensor (second proximity sensor) 7 is provided on the housing 4 in addition to the above-described configuration, and the proximity sensor (first proximity sensor) 6 is provided.
Before the fact that the face (mouth) of the speaker A has approached the microphone 3 is detected, it is detected that the speaker A has approached the apparatus (housing 4). The auxiliary proximity sensor (second proximity sensor) 7
The face (mouth) of the speaker A is detected by the proximity sensor (first proximity sensor) 6 according to the detection result by the microphone 3
Prior to detecting that the control unit 2 is approaching
, A recognition start command is given to the voice recognition unit 1 to start recognition processing.

The characteristic operation of this apparatus will be described with reference to FIG. 2. First, an auxiliary proximity sensor (second proximity sensor)
At 7, it is detected whether or not the utterer A has approached the present apparatus (housing 4) within a predetermined distance range, and the approach state is monitored (step a). When the auxiliary proximity sensor (second proximity sensor) 7 detects that the speaker A has approached the device (housing 4) within a predetermined distance range (for example, about 30 cm), the control unit 2 Outputs a recognition start command to the voice recognition unit 1 (step b), and causes the voice recognition unit 1 to start the voice recognition processing operation (step c).

In this state, the proximity sensor (first proximity sensor)
At 6, it is checked whether or not the face (mouth) of the speaker A has approached the microphone 3 within a predetermined distance range (for example, about 10 cm) (step d). The proximity detection of the speaker A's face (mouth) by the proximity sensor (first proximity sensor) 6 with respect to the microphone 3 is performed by the auxiliary proximity sensor (second proximity sensor) 7. Is performed while monitoring whether or not the face (mouth) of the speaker A has approached the microphone 3 within a predetermined time (step e). If the proximity is not detected within the predetermined time (time-out), A message to the effect that the face is brought close to the microphone is output using the display 5 (step f). This predetermined time is generally 2
It is sufficient to set the time to about 3 seconds.

When it is detected that the face of the speaker A has come sufficiently close to the microphone 3 through such a control procedure (step d), a message is output using the display 5 to prompt the user to start voice input. (Step g). The voice uttered in accordance with the message is input (step h), and the input voice is recognized by the voice recognition unit 1 (step i). Then, the recognition result is transferred from the voice recognition unit 1 to the control unit 2 (step j), and a confirmation display of the recognition result is appropriately performed to the speaker, thereby ending a series of voice recognition processing control. I do.

Thus, according to the present apparatus configured as described above and performing the control processing operation as described above, the recognition processing in the voice recognition unit 1 is started when the speaker A approaches, and thereafter, the speaker A Since a message prompting the start of voice input is output when approaching, even if the speaker A's utterance start timing is slightly earlier than the message display timing, the input voice can be input without any loss of the head part. Can be taken into the speech recognizing unit 1 to perform recognition output, and the voice uttered by the speaker A can be reliably recognized. Therefore, it is possible to eliminate the occurrence of erroneous recognition and recognition rejection due to the lack of the leading portion, and to improve the recognition efficiency. In addition, since the frequency of prompting the speaker A to re-input the voice is reduced as much as there is no misrecognition or recognition rejection due to the lack of the leading portion, the burden on the speaker can be reduced, and the apparatus itself has the same effect. It is possible to reduce the processing load.

Further, according to the embodiment, when a state in which the speaker A is not close to the microphone 3 is detected, it is possible to appropriately output a message for calling attention, thereby effectively stabilizing a voice input situation. It is possible to secure.

That is, the function is shared by the two proximity sensors 6 and 7, and the voice recognition process is controlled in accordance with the result of state detection by the proximity sensors 6 and 7, so that the voice generated by the speaker A is reliably recognized. It becomes possible.

Note that the present invention is not limited to the above-described embodiment. For example, by using a proximity sensor (distance measuring device) that can detect the proximity distance of the speaker in a stepwise manner in an analog manner, it is also possible for one sensor to perform substantially the two functions described above. Further, here, the structure in which the voice recognition unit 1 and the control unit 2 are housed in the housing 4 has been described. However, the microphone 3 and the proximity sensors 6, 7 and the housing may be provided separately. It is. Needless to say, the speech recognition system used here is not particularly limited, and can be variously modified and implemented without departing from the gist of the invention.

[Effects of the Invention] As described above, according to the present invention, the speech recognition processing is controlled so as to start the speech recognition processing before the speaker is detected to approach the microphone. Even in the early case, there is a great practical effect such that the uttered voice can be surely recognized without causing the loss of the head portion.

[Brief description of the drawings]

FIG. 1 shows a speech recognition apparatus according to one embodiment of the present invention. FIG. 1 is a schematic configuration diagram of a main part of the embodiment apparatus, and FIG. 2 is a diagram showing a control processing sequence thereof. 1 ... voice recognition unit, 2 ... control unit, 3 ... microphone, 4 ... housing, 5 ... display, 6 ... proximity sensor (first proximity sensor), 7 ... auxiliary proximity sensor (second 2 proximity sensor), A ... speaker.

Claims (1)

(57) [Claims] 1. A voice recognition device for recognizing an input voice by taking in a voice uttered by a speaker by a microphone provided in a housing, wherein a voice recognition device detects that the speaker has approached a first distance. 1
And a second sensor for detecting that the speaker has approached a second distance.
Means for outputting a message related to voice input to the speaker when the first proximity sensor detects that the face of the speaker has approached a first distance. Means for starting voice recognition for voice input from the microphone when the second proximity sensor detects that the speaker has approached a second distance. Speech recognition device.