JPH03212700A - signal processing device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Industrial applications The present invention relates to a signal processing device used for audio processing.

Prior Art FIG. 2 is a block diagram of a conventional signal processing device. As shown in the figure, the cepstrum calculating section 16 calculates the cepstrum of the audio input and supplies it to the peak detecting section 17.

The peak detection unit 17 is supplied with the cepstrum from the cepstrum calculation unit 16, detects its peak, and outputs the audio detection unit 19.
supply to. The audio detection section 19 receives the peak signal from the peak detection section 17, performs audio detection, and supplies the result to the matching section 2o. The speech analysis 18 analyzes the speech input and supplies it to the matching section 20 . The matching section 20 is configured to process the signal from the speech analysis section 18 in accordance with the signal supplied from the speech detection section 19 to obtain a recognition output.

The operation of the conventional signal processing device configured as described above will be described below. The cepstrum of the audio input is outputted by the cepstrum calculating section 16. Then, the peak detection section 17 detects the cepstrum peak. The voice detection section 19 determines the presence or absence of voice based on the presence or absence of the cepstrum peak or its magnitude, and supplies a control signal corresponding to the presence or absence of voice to the matching section 20. On the other hand, the audio input is analyzed by the audio analysis section 18 so that pattern matching can be performed by the matching section 20, and the obtained signal is supplied to the matching section 20. The matching section 20 performs speech recognition by matching the signal supplied from the speech analysis section 18 with a reference pattern according to the control signal from the speech detection section 19, and obtains a recognition output. Here, the control signal from the voice detection section 19 controls the matching section 20 to perform a matching operation whenever voice is detected.

Problems to be Solved by the Invention However, in such conventional signal processing devices, when a voice is input, the voice detection section is activated and a control signal is output, and the matching operation is always performed. The system also operates on voice inputs, resulting in unnecessary signal processing and unnecessary processing time, as well as the inconvenience that many misrecognitions occur.

The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to provide a signal processing device that operates only on voice input that is a recognition target.

Means for Solving the Problems In order to achieve the above object, the present invention provides a speech analysis section that analyzes speech input and outputs an analysis signal, and a matching section that compares the analysis signal with a template and outputs a recognition output. In the speech recognition means, the matching section is provided with a first control signal input means and a second control signal input means, a cepstrum calculation section that calculates from the speech input and outputs the cepstrum, and a cepstrum calculation section that calculates the peak of the cepstrum. a peak detection unit that detects and outputs a peak signal; a voice detection unit that determines the presence or absence of voice from the peak signal and outputs a first control signal; and a second control signal that corresponds to the mode setting input and the peak signal. and a peak value memory for storing the peak signal, and the control unit writes the peak signal in the memory in response to a mode setting input of "setting", and also outputs the control signal. The unit compares the peak signal of the memory and the cepstrum peak signal of the audio input in response to a mode setting input of “recognition °′, and outputs the second control signal in response to a difference in their quefrency; The matching section may be configured to output a recognition output in response to input of the first control signal and the second control signal.

Operation According to the present invention, with the above-described configuration, the cepstrum peak of the audio input is detected through the cepstrum section and the peak detection section. Then, the voice detection section determines the presence or absence of voice based on the peak of the cepstrum, and supplies a first control signal corresponding to the presence or absence of voice to the matching section. In addition, in the control section, when the mode setting input is "registration", the cepstrum peak signal obtained by the peak detection section is stored in the peak value memory, and when the mode setting input is "6 recognition", the peak detection section The cepstrum peak signal obtained in the above is compared with the peak value signal stored in the peak value memory, and a second control signal is supplied to the matching unit according to the difference in the respective quefrency.Furthermore, the audio input is processed by audio analysis. The matching section analyzes the data for use in the matching section, and performs the matching process with the data registered in advance in the matching section to obtain a recognition output.At this time, the start of the matching process starts with the first and a second control signal.That is, the first control signal from the voice detection unit is controlled by the second control signal from the control unit so as to start the matching operation when voice is detected. If the mode setting input is "recognition", it is determined that there is no difference between the quefrency of the cepstrum peak signal of the audio input and the quefrency value of the peak signal registered in advance in the memory when the mode setting is "setting". start the matching operation.

EXAMPLE Hereinafter, an example of the present invention will be described with reference to FIG.

FIG. 1 is a block diagram of a signal processing device in one embodiment of the present invention. In the figure, a cepstrum calculating section 1 calculates a cepstrum from a voice input and supplies it to a peak detecting section 2. The peak detection section 6 detects peaks from the cepstrum and supplies them to the control section 3 and the voice detection section 6, respectively.

The voice detection section 6 detects the presence or absence of voice based on the presence or absence of the cepstrum peak signal supplied from the peak detection section 2, and supplies a first control signal to the matching section 7. The control unit 3 stores the cepstrum peak signal supplied from the peak detection unit 2 in the peak value memory 4 according to the mode setting input.
A second control signal is output to the matching unit 7 using the data supplied to the peak value memory 4 or the data supplied from the peak value memory 4 . The peak value memory 4 stores the cepstrum peak signal of the peak detection section 2, and stores the data via the control section 3.

Perform reading. The speech analysis section 6 analyzes the speech input into a data format used by the matching section 7, and supplies an analysis signal to the matching section 7. The matching section 7 includes the speech analysis section 5
The analysis signal analyzed by the voice detection section 6 and the control section 3
The first and second control signals are supplied from the speech analyzer 6, and in accordance with the control signals, the analyzed signal supplied from the speech analyzer 6 is compared with a template to obtain a recognition output.

The operation in the above configuration will be explained. First, when the mode setting input is "Register", the cepstrum of the audio input is calculated in the cepstrum calculation unit 1, the cepstrum peak is detected in the peak detection unit 2, and is supplied to the control unit 3. and stored in the peak value memory 4. Then, the control section 3 sends a second control signal to the matching section 7 for not performing the matching process. Similarly, when the mode setting input is "recognition", the cepstrum of the voice input is calculated in the cepstrum calculation section 1, and the cepstrum peak is detected in the peak detection section 2. The voice detection unit 6 determines the presence or absence of voice depending on the presence or absence of the cepstrum peak signal. If it is voice, it sends a first control signal to perform the matching process, and if it is not voice, it sends a first control signal to not perform the matching process. 1 control signal is sent to the matching section 7.

At the same time, the cepstrum peak signal of the peak detection section 2 is compared with the content of the peak value memory 4 stored in advance by the control section 3, and if the quefrency values of both are close, the cepstrum peak signal of the peak detection section 2 is compared with the content of the peak value memory 4 stored in advance. Second
If the two quefrency values are not close to each other, a second control signal is sent to the matching unit 7 for not performing the hemanotching process. Then, in the matching section 7, the analysis signal analyzed by the speech analysis section 5 is
When the first and second control signals supplied from the voice detection section 6 and the control section 3 are both signals in which matching processing is to be performed, the data is compared with the template data, recognition processing operation is performed, and the result is Output as recognition output. As described above, according to the signal processing device of the embodiment of the present invention,
Matching processing with the template is performed only when the cepstrum peak of the audio input, that is, the pitch frequency of the speaker, is close to the one registered in advance, so matching processing is not performed when the audio input is from a speaker other than the registered speaker. , the processing time required for the matching process of the matching unit can be eliminated, and when a voice input from a person other than the registered speaker is input, a rejection result is immediately output.

Furthermore, if the device is configured with a microprocessor, etc.
Since the matching process can be minimized, the load on the CPU can be reduced and the load can be used for other processes.

It goes without saying that outputting a recognition result indicating that the speaker is different from the registered speaker can be easily achieved by using the control signal from the control section 30.

Effects of the Invention As is clear from the embodiments described above, the present invention is capable of storing the peak signal output of the cepstrum peak detector in the peak value memory or storing the peak signal output of the peak detector as the peak value, depending on the mode setting input. By having a control section that supplies a second control signal to the matching section compared to the memory, it is possible to perform the matching operation only when the pinch frequency of the audio input is close to the pre-registered one. When the voice of a person other than the speaker is input, matching processing is not performed, and this processing process can be omitted, and a rejection result can be obtained at high speed. Furthermore, when the device is configured with a microprocessor, it is possible to reduce the matching processing process to the necessary minimum, and the load on the CPU can be greatly reduced, which can be used for other processing processes. This has the effect of streamlining the design.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a signal processing device according to an embodiment of the present invention, and FIG. 2 is a block diagram of a conventional signal processing device. 1...Cepstrum calculation unit, 2...Peak detection unit, 3...Control unit, 4...Peak value memory, 5...Speech analysis unit , 6... Voice detection section, 7... Matching section.

Claims (1)

[Claims] a voice analysis unit that analyzes voice input and outputs an analysis signal;
In a speech recognition means comprising a matching section that compares the analysis signal with a template and outputs a recognition output, the matching section is provided with a first control signal input means and a second control signal input means, and calculation is performed from the speech input. a cepstrum calculation unit that outputs a cepstrum; a peak detection unit that detects a peak of the cepstrum and outputs a peak signal; and a voice detection unit that determines the presence or absence of voice from the peak signal and outputs a first control signal. a control unit that outputs a second control signal in response to the mode setting input and the peak signal; and a peak value memory that stores the peak signal; Correspondingly, the peak signal is written in the memory, and the control unit compares the peak signal in the memory with the cepstral peak signal of the audio input in response to the mode setting input being "recognition", and determines the respective quefrency. The signal processing device outputs the second control signal in response to a difference, and the matching unit outputs a recognition output in response to input of the first control signal and the second control signal.