JPH0559500U - Voice recognizer - Google Patents

Abstract

(57) [Abstract] [Purpose] The noise data of the voice pattern to be referred to during voice recognition is fixed. [Structure] A voice signal detected by a microphone 1 is band-limited and amplified by a preamplifier 2 having a bandpass function, and input to a voice feature extraction unit 3. The voice feature extraction unit 3 converts the input voice analog signal into a voice digital signal, extracts the feature, and outputs the feature to the recognition processing unit 4. The recognition processing unit 4 operates according to the program of the software recording ROM 5, and at the time of voice registration, a voice pattern containing no noise is stored in the RAM 6 for the first time, and stored in the noise pattern recording ROM 10 for the second time. The present data is automatically taken out and the voice pattern including noise is stored in the RAM 7 together with the data. Recognition processing unit 4
Also recognizes a voice by referring to the data in the RAMs 6 and 7 from the signal output from the voice feature extraction unit 3 during voice recognition.

Description

[Detailed description of the device]

[0001]

[Industrial applications]

 The present invention relates to a voice recognition device, and more particularly to a voice recognition device that prevents erroneous recognition due to noise when voice recognition is performed in a noisy environment.

[0002]

[Prior Art]

 Conventionally, as a practical problem of speech recognition, dealing with noise has been the biggest problem. Although various attempts have been made to prevent erroneous recognition due to noise, one method is to use a voice pattern that is quietly recorded as a voice pattern to be referred to during voice recognition and a voice pattern that is recorded with noise. There is a method that uses.

FIG. 3 shows a speech recognition apparatus of such a system. The voice signal detected by the microphone 1 is band-limited and amplified by the preamplifier 2 having a bandpass function, and input to the voice characteristic extraction unit 3.

The voice feature extraction unit 3 converts the input voice analog signal into a voice digital signal, extracts the feature, and outputs the feature to the recognition processing unit 4. The recognition processing unit 4 operates according to the program of the software recording ROM 5, and at the time of voice registration, stores a voice pattern containing no noise in the RAM 6 and a voice pattern containing noise in the RAM 7. The recognition processing unit 4 also recognizes a voice by referring to the data in the RAMs 6 and 7 from the signal output from the voice feature extraction unit 3 during voice recognition, outputs a signal to the signal generation unit 8, and outputs the signal to the signal generation unit. 8 to operate the car audio device 9.

FIG. 4A shows a noise-free speech spectrum stored in the RAM 6, and FIG. 4B shows a noise-containing speech spectrum stored in the RAM 7. By referring to the two voice spectra in this way, erroneous recognition due to noise can be prevented.

[0006]

[Problems to be solved by the device]

 In the above-mentioned conventional voice recognition device, the voice is input in a quiet environment during the first voice registration, but in the noise environment during the second voice registration (for example, the engine is turned on and the air conditioner is turned on). The voice is input with. This noise condition is difficult to be constant, and for example, there is a problem that a desired voice pattern may not be registered due to a change in the noise condition when the vehicle runs.

The present invention has been made in view of the above points, and an object of the present invention is to provide a voice recognition device which can always obtain constant noise data.

[0008]

[Means for Solving the Problems]

 The speech recognition device of the present invention registers a voice pattern recorded in a noise-free state and a voice pattern including noise recorded together with noise in a RAM, and stores two voice patterns registered in the RAM. When a noise is recorded in advance, the pattern is stored in ROM, and the pattern of the sound including the noise is registered in the RAM in the voice recognition device in which the erroneous recognition due to the noise is prevented by recognizing the sound by referring. In addition, the noise pattern data stored in the ROM is used.

In addition, when the voice pattern is registered in the RAM a plurality of times in the voice recognition device, the noise pattern data stored in the ROM is automatically recalled when the voice pattern including noise is registered. It is configured to.

[0010]

[Action]

 According to the voice recognition device of the present invention, when the voice pattern is registered in a noisy environment, since the data of the ROM in which the standard noise pattern is stored in advance is used, a constant noise data is always registered and the noise condition changes. False recognition due to is eliminated.

[0011]

【Example】

 An embodiment of this invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a voice recognition device according to an embodiment of the invention. The voice signal detected by the microphone 1 is band-limited and amplified by the preamplifier 2 having a bandpass function and input to the voice feature extraction unit 3.

The voice feature extraction unit 3 converts the input voice analog signal into a voice digital signal, extracts the feature, and outputs the feature to the recognition processing unit 4. The recognition processing unit 4 operates according to the program of the software recording ROM 5, and at the time of voice registration, as shown in FIG. 2, the first time, the RAM 6 stores a voice pattern containing no noise, and the second time, the noise pattern is stored. The noise data stored in advance in the recording ROM 10 is automatically taken out, and the voice pattern including the noise is stored in the RAM 7 together with the data. The recognition processing unit 4 also recognizes a voice by referring to the data of the RAMs 6 and 7 from the signal output from the voice feature extraction unit 3 during voice recognition, and outputs the signal to the signal generation unit 8 to generate the signal. The section 8 operates the car audio device 9.

By thus referring to the two voice spectra, erroneous recognition due to noise can be prevented. Also, since the noise data is constant, it is not necessary to consider the environment during noise recording.

[0014]

[Effect of the device]

 According to the voice recognition device of the present invention, it is not necessary to generate noise at the time of voice registration, and therefore input is simple.

Further, since there is no variation in noise and the noise is constant, erroneous recognition due to changes in noise is eliminated. Also, when registering a voice, the noise mode is automatically set at the time of the second recording, and the operation is easy.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a voice recognition device according to an embodiment of the invention.

FIG. 2 is a graph showing a voice signal at the time of voice registration of the voice recognition device.

FIG. 3 is a block diagram showing an example of a configuration of a conventional voice recognition device.

FIG. 4 is a graph showing a spectrum of a registered voice in the voice recognition device.

[Explanation of symbols]

 1 Microphone 2 Preamplifier 3 Voice feature extraction unit 4 Recognition processing unit 5 Software recording ROM 6 RAM 7 RAM 8 Signal generating unit 9 Car audio device 10 Noise pattern recording ROM

Claims (2)

[Scope of utility model registration request] 1. A voice pattern recorded without noise and a voice pattern containing noise recorded together with noise are registered in a RAM, and the voice is referred to by referring to the two voice patterns registered in these RAMs. In a voice recognition device in which erroneous recognition due to noise is prevented by recognizing, noise is recorded in advance, its pattern is stored in a ROM, and the pattern of the voice including the noise is registered in the RAM.
A voice recognition device characterized by using noise pattern data stored in M. 2. When the voice pattern is registered in the RAM a plurality of times, the R is added when the voice pattern including noise is registered.
The voice recognition device according to claim 1, wherein the noise pattern data stored in the OM is automatically called.