JP2754960B2 - Voice recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition apparatus capable of controlling an external device by using speech, and more particularly, to erroneous speech recognition in a situation where noise or noise is large. The present invention relates to a speech recognition device capable of preventing an error in control of an external device in a situation.

[0002]

2. Description of the Related Art A variety of speech recognition devices for recognizing speech and controlling external devices have been put to practical use. In such a device, there is a problem that if all sounds or voices are to be recognized, they are easily affected by noise or the like. Therefore, conventionally, a certain type of speech recognition apparatus adopts a method in which a threshold for detecting speech is set high, and speech recognition is performed only when a large speech is input to control an external device. In other voice recognition devices, the threshold value for voice detection is set to a normal level, and the similarity determination in voice recognition is made stricter until words such as predetermined command words are recognized. In order to reduce false recognition, a method of controlling an external device has been adopted.

[0003]

However, in the former of such conventional speech recognition apparatuses, when a large noise exceeding a threshold value for speech detection is input, an external apparatus operates. was there. Also, in the latter device, since the sound level is not a problem, noise or conversations occurring around the device may be erroneously recognized as a command word.In such a case, the user controls the external device. There is a disadvantage that the external device may malfunction even when it is not desired.

It is an object of the present invention to provide a speech recognition device that can prevent erroneous control of an external device due to erroneous recognition.

[0005]

According to the present invention, there is provided a voice detecting section for detecting voice, and the voice detecting section detects voice.
A minimum audio level setting means for setting a minimum audio level in conjunction with a noise signal level immediately before the time point, and a maximum level of the audio when the audio detection unit ends detection of the audio.
Determining a level comparing means for comparing whether greater than the minimum sound level, the degree of similarity between the voice standard pattern registered in advance and the sound when the maximum level of the voice is determined to be greater than the minimum sound level The speech recognition device includes a similarity determination unit and a recognition result output unit that outputs a similarity determination result as a speech recognition result.

That is, according to the first aspect of the present invention, the minimum sound level is set in conjunction with the noise signal level immediately before the sound detection unit detects the sound, and when the sound is detected, the maximum level in the sound is minimized. Only when the voice level is higher than the voice level, the similarity with the voice standard pattern is determined, so that erroneous control of the external device due to erroneous recognition can be prevented. Furthermore, in the invention described in claim 2,
A speech for analyzing and encoding the speech by the speech recognition device according to 1.
The analysis unit encodes the speech to be recognized by the device
Storage unit for storing as a standard pattern to determine similarity
The means is based on the voice standard pattern stored in this storage
The similarity is compared by comparing with the speech coded by the speech analysis unit.
It is characterized by determining. Further according to claim 3
According to the present invention, in the speech recognition apparatus according to claim 1, the minimum speech
When the minimum sound level is set by the bell setting means
The voice standard pattern registered during voice detection and this voice
A similarity calculating means for calculating the similarity with the
The determining means determines the maximum level of the sound by the level comparing means.
Is determined to be greater than the minimum sound level
The similarity calculated by the similarity calculating means is set in advance.
Determines whether a similarity higher than the similarity determination value has been obtained
It is characterized by that. That is, the invention of claim 3
In the figure, the noise signal immediately before the sound detector detects the sound
The minimum audio level is set in conjunction with the
If detected, the voice and the pre-registered voice standard pattern
Is calculated. And voice detection
Is completed and the maximum level of the audio is
Similar to this audio standard pattern only if it is larger than
By comparing the degree with a preset similarity judgment value,
Determines whether the detected voice is similar to the standard pattern.
Erroneous control of external devices due to erroneous recognition.
Can be prevented.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to embodiments.

FIG. 1 shows an outline of the configuration of a speech recognition apparatus according to an embodiment of the present invention. An audio signal 12 output from a microphone 11 that captures a speaker's audio
Is input to the voice detection unit 13. The level monitoring unit 14
The input signal level of the audio signal 12 is constantly monitored. The detected audio signal 15 is input to an audio analysis unit 16 and encoded to extract an audio pattern 17. The extracted voice pattern 17 is input to the similarity calculation unit 18. The similarity calculation unit 18 stores the standard pattern 2 stored in the storage unit 19.
1 is used to calculate the degree of similarity with the audio pattern 17. Here, the standard pattern 21 is obtained by encoding numbers and the like in addition to the command voice.

The result 22 of the similarity calculated by the similarity calculating section 18 is sent to a similarity determining section 23. The similarity determination unit 23 determines that the voice pattern 17 is smaller than a preset value.
Is high, it is determined that the speaker's voice matches the standard pattern 21. In this case, the recognition result 24 is sent to the recognition result sending unit 25. The recognition result sending unit 25 outputs a control signal 26 for controlling an external device (not shown) based on the recognition result sending unit 25. The level monitoring unit 14, the voice analysis unit 16, the similarity calculation unit 18, the storage unit 19, the similarity determination unit 23, and the recognition result sending unit 25 described above are connected to the control unit 27, and control is performed by the control unit 27. It has become.

FIG. 2 specifically shows the level monitor shown in FIG. The level monitoring unit 14 includes a level monitoring circuit 31 connected to the audio detection unit 13, a level storage circuit 32, and a level comparison circuit 33. The level monitoring circuit 31 constantly monitors the input signal level of the audio detection unit 13. This level monitoring circuit 3
The level storage circuit 32 connected to 1 sets a minimum audio level linked to the level of the noise signal of the audio detection when the audio detection is started, and stores this. The level comparison circuit 33 compares the audio signal level detected by the level monitoring circuit 31 during audio detection with the minimum audio level stored in the level storage circuit 32. Then, at the end of the voice detection, if the maximum level of the voice signal during the voice detection exceeds the minimum voice level, a signal for causing the control unit 27 to perform the similarity determination is transmitted. . In addition,
The level monitor 14 is controlled by a CPU (Central Processing Unit) (not shown) together with other circuit parts in the speech recognition device shown in FIG. A program for this control is stored in a ROM (read only memory) not shown.

The operation of the speech recognition apparatus configured as described above will be specifically described with reference to FIG. It is assumed that the standard pattern of the control voice stored in the storage unit 19 of the voice recognition device shown in FIG. When the control operation of the voice recognition is started, a state of waiting for voice detection is entered as shown in FIG. 3 (step S101). In this state, the voice detection unit 13 constantly monitors the signal from the microphone 11. When the voice detection is started (Y), the level monitoring circuit 31 shown in FIG. 2 monitors the signal level of the voice analysis unit 16 and sets the minimum voice level linked to the level of the noise signal immediately before the voice detection starts. (Step S102). That is, when the speaker utters “○○○” and a temporally leading portion of the voice is detected, the minimum voice level is set at the same time. This minimum audio level is stored in the level storage circuit 32 of FIG.

When the minimum audio level is set, the level monitoring circuit 31 monitors the audio level, and the level comparison circuit 33 outputs the audio signal level detected by the level monitoring circuit 31 during the audio detection and the level stored in the level storage circuit 32. Compare with the stored minimum audio level. At the same time, the similarity between the standard pattern stored in the storage unit 19 and the input voice pattern is calculated by the similarity calculation unit 18 (step S10).
3).

The monitoring of the voice level and the calculation of the similarity between the voice pattern are performed until the voice detection is completed (step S104). When the detection of the voice is completed (Y), the level comparison circuit 33 determines whether or not the maximum level of the voice in “○○○” uttered by the speaker is higher than the above-described minimum voice level (step S105). As a result, when it is determined that it is equal to or lower than the minimum voice level (N), the similarity determination unit 23 does not perform similarity determination, and returns to step S101 again to return to the next voice. It will be waiting for input.

On the other hand, "○○○" which the speaker uttered
If the maximum level of the voice is higher than the minimum voice level (step S105; Y), the similarity determination unit 23 determines the similarity. If it is determined that the speaker has produced a sound that is not similar to the registered standard pattern (step S106; N), step S10 is performed.
The process returns to step 1 to wait for the next voice input.

On the other hand, when a similarity higher than a predetermined similarity determination value is obtained, such as when the speaker utters a voice “Kidou”, it is determined that the pattern is similar to the standard pattern. (Step S106; Y). in this case,
Upon receiving the recognition result 24, the recognition result sending unit 25 outputs a control signal 26 corresponding to the recognition result to the external device (step S107). Thus, a series of operations is completed (END).

[0016]

As described above, according to the present invention, the minimum sound level is set in conjunction with the noise level immediately before the time when sound is detected, and the sound standard pattern is set only when a sound signal exceeding this level is input. Therefore, even in a noisy environment or when noise is generated due to surrounding conversation, erroneous recognition due to erroneous voice detection can be prevented, and the external device can be identified. There is an effect that erroneous control can be surely prevented. Further claims
According to the invention described in (3), immediately before the time when the voice is detected,
Set the minimum audio level in conjunction with the noise level
Similarity with a pre-registered voice standard pattern during voice detection
Was calculated. When voice detection is completed,
When the maximum level of the sound exceeds the minimum level
Only this similarity is compared with a preset similarity determination value.
To determine if they are similar to the audio standard pattern
After the voice detection has been completed,
So that other voice detection is performed during voice recognition.
Will be able to do it.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an outline of a configuration of a speech recognition device according to an embodiment of the present invention.

FIG. 2 is a block diagram specifically showing a level monitoring unit shown in FIG.

FIG. 3 is a flowchart showing a state of control of the voice recognition device of the embodiment.

[Explanation of symbols]

 Reference Signs List 11 microphone 13 voice detection unit 14 level monitoring unit 16 voice analysis unit 18 similarity calculation unit 19 storage unit 23 similarity determination unit 25 recognition result transmission unit 27 control unit 31 level monitoring circuit 32 level storage circuit 33 level comparison circuit

Claims (3)

(57) [Claims] An audio detection unit for detecting audio; a minimum audio level setting means for setting a minimum audio level in conjunction with a noise signal level immediately before the audio detection unit detects the audio; Level comparing means for comparing whether or not the maximum level of the voice is higher than the minimum voice level when the detection unit has finished detecting the voice; and determining that the maximum level of the voice is higher than the minimum voice level. And a recognition result output unit that outputs a similarity determination result as a speech recognition result. Voice recognition device. 2. A speech analysis unit for analyzing and encoding speech, and a storage unit for encoding speech to be recognized by the device and storing the encoded speech as a speech standard pattern, wherein the similarity determination means includes 2. The speech recognition apparatus according to claim 1, wherein a similarity is determined by comparing the stored speech standard pattern with speech encoded by the speech analysis unit. 3. The minimum audio level setting means,
When a low sound level is set, it is registered in advance during sound detection.
Class for calculating the similarity between a selected audio standard pattern and this audio
A similarity calculating means, The similarity determination unit outputs a sound by the level comparison unit.
The maximum voice level is determined to be greater than the minimum voice level.
Similarity calculated by this similarity calculation means
Obtained a higher similarity than the preset similarity judgment value.
2. The voice according to claim 1, wherein it is determined whether or not the voice is received.
Recognition device.