JP2003271175A - Device and program for evaluating acoustic processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a sound processing evaluating device capable of easily performing evaluation under various conditions. <P>SOLUTION: This acoustic processing method evaluating device is provided with a data storing means 1 for previously storing a voice signal recorded with voices and data such as the transfer characteristic of a sound in an automobile, an acoustic data working means 2 for working a plurality of pieces of data into acoustic data for evaluation, an acoustic processing means 3 for processing the acoustic data for evaluation on the basis of an acoustic processing method of an evaluation object such as voice recognition processing, and an evaluation outputting means 4 for performing evaluation on the basis of a result processed by the acoustic processing means and outputting the evaluated result. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention prepares acoustic data for evaluation of various evaluation conditions based on the data stored in the data storage means as a database, and evaluates the acoustic processing method prepared in advance. The present invention relates to a sound processing method evaluation device and a sound processing method evaluation program.

[0002]

2. Description of the Related Art Generally, when a computer executes a speech recognition processing program and comparatively evaluates a processing method in which acoustic data is speech-recognized from a processing result, as evaluation acoustic data for each evaluation condition for evaluating the processing method. For example, sound data for evaluation, such as voice data uttered in a car, voice data uttered in an office, and voice data uttered when music is playing, have been prepared.
Further, as an evaluation device for preliminarily storing digital signal data of a specific pattern for evaluating the sound processing method, the one described in JP-A-7-287595 is known and is shown in FIG.

The conventional acoustic processing system evaluation apparatus shown in FIG. 14 has a test signal generating section 38 for generating a test digital signal and an audio signal synthesized from the test signal by the speech synthesizing section 20 of the speech processing apparatus 1. The test signal generation section 38 stores a digital pattern data of a specific pattern for voice synthesis section evaluation in advance in a storage section 31 such as a ROM and a storage section 31.
The digital signal generating circuit 32 for generating a digital signal based on the test data stored in the memory 32 and the interface 33 for outputting the digital signal to the outside, and the evaluation unit 39 outputs the audio signal output from the voice synthesis unit 20. It is composed of a conversion circuit 34 for converting a signal into an effective value, a sampling circuit 35 for sampling a voltage level converted into an effective value, and a CPU 36 for judging pass / fail of the voice synthesizer 20 from the sampled data. .

Next, the function of each part will be specifically described. The storage unit 31 is a memory in which digital signal data of a specific pattern for voice synthesis unit evaluation is stored in advance in order to evaluate the voice synthesis unit 20 of the voice processing device 1. The stored test data is a voice Speech analysis unit 1 in the processing device 1
The digital signal data has the same format as the digital signal protocol output from 0.

The digital signal generation circuit 32 is a circuit for generating a digital signal based on the test data stored in the storage unit 31 when the voice synthesis unit 20 is evaluated, and the interface 33 is operated from the digital signal generation circuit 32. The generated digital signal is output to the outside.

The RMS conversion circuit 34 is a circuit for taking the root mean square of the analog value of the input audio signal over one cycle of the voltage value and outputting the square root of the mean of the square as the effective value. An analog signal obtained by synthesizing a test signal generated by the digital signal generating circuit 32 based on the test data stored in 31 by the voice synthesizing unit 20 in the voice processing device 1 and converting into an audio signal is input and its effect is obtained. It is designed to output the value.

The sampling circuit 35 samples the effective value data of the audio signal output from the RMS conversion circuit 34 at a constant sampling interval and converts it into digital data, which is converted into a serial or parallel signal and is sent to the CPU.
It is a circuit which outputs to 36.

The CPU 36 takes in the digital data output from the sampling circuit 35, and the audio processor 1
It is a determination unit that determines the quality of the voice synthesis unit 20 inside.
In addition, as a quality determination method, the result of a simulation for synthesizing a voice from digital signal data of a specific pattern for evaluation stored in advance in the storage unit 31 is incorporated into the evaluation program to obtain a simulation result. The voice synthesizer 2 compares the data actually input from the sampling circuit 39 and determines whether or not a close value is obtained.
The quality of 0 is determined.

[0009]

However, in such a conventional acoustic processing method evaluation apparatus, the more conditions for evaluating the acoustic processing method increase, the more the amount of evaluation data prepared. Therefore, there are problems that it takes time and effort to prepare the evaluation data and the storage area of the evaluation data of the evaluation system increases.

The present invention has been made to solve such a problem, and is stored in the data storage means as a database without preparing a huge amount of acoustic data for evaluation for evaluating the acoustic processing method. An acoustic processing method evaluation device and an acoustic processing method evaluation program that create acoustic data for evaluation of various evaluation conditions based on the data and evaluate the result of processing the generated acoustic data for evaluation by the acoustic processing method. It is provided.

[0011]

A sound processing method evaluation apparatus according to the present invention comprises a data storage means for storing data in advance, a sound data processing means for processing a plurality of the data into sound data for evaluation, and an evaluation target. Sound processing means for processing the sound data for evaluation based on the sound processing method described above, and evaluation output means for evaluating based on the result processed by the sound processing means and outputting the evaluated result. Have a configuration. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

The data storage means has a data input section for inputting data. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic data processing means has a processing content input section for inputting a processing content for processing the data into acoustic data for evaluation. With this configuration,
Sound data for evaluation of various evaluation conditions were created based on the data stored in the data storage means as a database, without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The processing result in which the evaluation audio data is processed by the audio processing method is evaluated.

Further, the sound processing means has a sound processing method input section for inputting a sound processing method for processing the sound data. With this configuration, without preparing a huge amount of acoustic data for evaluating the acoustic processing method,
To create acoustic data for evaluation of various evaluation conditions based on the data stored in the data storage means as a database, and to evaluate the processing result obtained by processing the created acoustic data for evaluation by the acoustic processing method. Become.

Further, the sound processing system evaluation apparatus of the present invention comprises a data storage means for pre-storing data, an acoustic data processing means for processing a plurality of the data into evaluation acoustic data, and the data for the evaluation. Weighting means for weighting the evaluation acoustic data for evaluation based on the processing content processed into the acoustic data, and acoustic processing for processing the evaluation acoustic data based on the evaluation target acoustic processing method. Means, and an evaluation output means for evaluating based on a result processed by the sound processing means and a weight attached to the processed acoustic data for evaluation, and outputting an evaluation result. Have With this configuration, it is possible to create and create acoustic data for evaluation of various evaluation conditions based on the data stored in the data storage means as a database without preparing a large amount of evaluation data for evaluating the acoustic processing method. The processing result obtained by processing the evaluated acoustic data by the acoustic processing method is evaluated.

Further, the weighting means has a weight input section for inputting an evaluation weight to the evaluation acoustic data based on a processing content in which the data is processed into the evaluation acoustic data. . With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic data processing means includes a transfer characteristic convolution unit for convolving transfer characteristics with the voice data stored in the data storage means, and noise data for the voice data convoluted with the transfer characteristics. It has a noise addition unit for adding. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic data processing means includes a first transfer characteristic convolution unit for convolving transfer characteristics with the voice data stored in the data storage means, and transfer characteristics with the noise data stored in the data storage means. A second transfer characteristic convolution unit that convolves the following, an amplitude multiplication unit that multiplies the convolved noise data by the amplitude, and a noise addition unit that adds the convoluted voice data and the amplitude-multiplied noise data. Have With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic processing means executes a voice recognition process prepared in advance on the acoustic data for evaluation,
The evaluation output means evaluates the voice recognition processing based on the result processed by the acoustic processing means. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic processing means executes speaker authentication processing prepared in advance for the acoustic data for evaluation,
The evaluation output means evaluates the voice recognition processing based on the result processed by the acoustic processing means. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic processing means executes a noise suppression process prepared in advance on the acoustic data for evaluation,
The evaluation output means evaluates the noise suppression processing based on the result processed by the acoustic processing means. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the acoustic processing means executes a voice detection process prepared in advance on the acoustic data for evaluation,
The evaluation output means evaluates the voice detection processing based on the result processed by the sound processing means. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, an acoustic data storage step of pre-storing data in a computer, an acoustic data processing step of processing a plurality of the data into acoustic data for evaluation, and the evaluation data based on an acoustic processing method to be evaluated The acoustic processing step of processing the acoustic data of 1. and the evaluation output step of evaluating based on the result processed in the acoustic processing step and outputting the evaluated result are executed. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer is made to execute a data input step of inputting the data. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer is made to execute a processing content inputting step of inputting processing details for processing the data into the acoustic data for evaluation. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer is made to execute an acoustic processing method input step of inputting a processing method for processing the evaluation acoustic data. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Furthermore, an acoustic data storing step of pre-storing data in a computer, an acoustic data processing step of processing a plurality of the data into evaluation acoustic data, and the data are processed into the evaluation acoustic data. A weighting step of weighting the evaluation acoustic data on the basis of the processed contents, an acoustic processing step of processing the evaluation acoustic data on the basis of the acoustic processing method to be evaluated, and the acoustic processing An evaluation output step of outputting an evaluation based on a result processed in the step and a weight attached to the processed acoustic data for evaluation is executed. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer is caused to execute a weight input step of inputting an evaluation weight to the evaluation acoustic data based on a processing content in which the data is processed into the evaluation acoustic data. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer is made to execute a transfer characteristic convolution step of convolving transfer characteristics with voice data and a noise addition step of adding noise data to the voice data with the transfer characteristics convoluted. There is. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, a computer performs a first transfer characteristic convolution step of convolving transfer characteristics with voice data, a second transfer characteristic convolution step of convolving transfer characteristics with noise data, and an amplitude with respect to the convoluted noise data. The amplitude multiplication step of multiplying by and the noise addition step of adding the convoluted voice data and the noise data multiplied by the amplitude are executed. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer performs a voice recognition process prepared in advance on the acoustic data for evaluation in the acoustic processing step, and based on the result processed in the acoustic processing step in the evaluation output step. The voice recognition process is evaluated. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer executes the speaker authentication process prepared in advance in the acoustic data for evaluation in the acoustic processing step, and based on the result processed in the acoustic processing step in the evaluation output step. The voice recognition process is evaluated. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer performs a noise suppression process prepared in advance on the acoustic data for evaluation in the acoustic processing step, and based on the result processed in the acoustic processing step in the evaluation output step. The noise suppression processing is evaluated. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

Further, the computer performs a voice detection process prepared in advance on the acoustic data for evaluation in the acoustic processing step, and based on the result processed in the acoustic processing step in the evaluation output step. The voice detection processing is evaluated. With this configuration, sound data for evaluation of various evaluation conditions can be created based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. The created acoustic data for evaluation is used to evaluate the processing result processed by the audio processing method.

[0035]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

(Embodiment 1) As shown in FIG. 1, an acoustic processing method evaluation apparatus according to the present embodiment includes a data storage means 1 for storing data and a plurality of the data processed into evaluation acoustic data. Sound data processing means 2, a sound processing means 3 for processing the sound data for evaluation based on a sound processing method to be evaluated, and an evaluation based on a result processed by the sound processing means 3, and this evaluation The evaluation output means 4 for outputting the result is provided.

The data stored in advance includes, for example, a voice signal in which voice is recorded, an acoustic signal in which music is recorded, a noise signal in which ambient noise heard on the road is recorded, and a sound in a car. Transfer characteristics are included. Also,
Processing the acoustic data for evaluation based on the acoustic processing method of the evaluation target means, for example, when performing speech recognition processing, a speech recognition processing program of the speech recognition processing method of the evaluation target to be evaluated is prepared in advance in the acoustic processing means. The acoustic processing means is to execute the speech recognition of the acoustic data for evaluation according to the flow of the speech recognition processing program of the speech recognition processing method of the evaluation target prepared in advance.

Next, as a specific example, referring to FIGS. 2 and 13, a voice uttered in an automobile is processed, and the processed voice is subjected to voice recognition processing by a voice recognition processing method prepared in advance. The operation of the acoustic processing method evaluation apparatus according to the present embodiment will be described according to the flow of evaluating a preset speech recognition processing method from the result of the recognition processing.

First, in the data storing step S1, the data storing means 1 stores a voice database 11 in which voices uttered by a speaker are stored, a transfer characteristic database 12 in which transfer characteristic data of a vocal environment is stored, and a vocal environment. The data is stored in advance in the noise database 13 in which the noise data added in is stored. Next, acoustic data processing step S2
In the transfer characteristic convolution step S21, the transfer characteristic convolution unit 21 of the acoustic data processing unit 2 convolves the transfer characteristic data in the automobile, which is the utterance environment, with the voice data. Next, in the noise addition step S22 of the acoustic data processing step S2, the noise addition section 22 of the acoustic data processing means 2 adds the in-vehicle noise data to the evaluation voice data in which the transfer characteristic is convoluted, and the evaluation sound is added. It is output as data. Next, in the acoustic processing step S3, the acoustic processing means 3 executes the speech recognition processing of the acoustic processing method of the evaluation target prepared in advance. Finally, in the evaluation output step S4,
The evaluation output unit 4 evaluates the voice recognition method to be evaluated based on the processing result of the voice recognition.

Here, the voice recognition processing is shown as an example, but it is also possible to evaluate the respective processing methods by executing speaker authentication processing, noise suppression processing, and voice detection processing.

Therefore, since the sound data, transfer characteristic data, and noise data are classified and stored in the data storage means 1 as a database, it is not necessary to have the same acoustic data for evaluation as the evaluation conditions of the acoustic processing method. It becomes possible to create acoustic data of various evaluation conditions only by having transfer characteristic data and noise data corresponding to the evaluation conditions. Note that, here, an example in which the voice recognition processing is performed by the acoustic processing unit 3 and the transfer characteristic convolution unit 21 and the noise addition unit 22 are illustrated as the calculation example of the acoustic data processing unit 2 has been described. The calculation processed into the acoustic data for evaluation can be changed according to the evaluation condition. For example, as shown in FIG. 3, acoustic data processing means 2
The processing process by may include the first convolution unit 21a, the second convolution unit 21b, the noise addition unit 22, and the amplitude multiplication unit 23. With this configuration, in the first convolution step S21a, the transfer characteristic stored in the transfer characteristic database 12 is convolved with the voice data stored in the voice database 11 by the first convolution unit 21a. Next, in the second convolution step S21a, the transfer characteristic stored in the transfer characteristic database 12 is convolved with the noise data stored in the noise database 13 by the second convolution unit 21b. Next, in the amplitude multiplication step S23, the amplitude multiplication unit 23 multiplies the amplitude component included in the noise data in which the transfer characteristic is convolved by a constant number. Finally, in the noise adding step S22, the noise adding unit 22 adds the noise data having the transfer characteristics convoluted to the sound data having the transfer characteristics convoluted, and outputs the result as evaluation data to the acoustic processing means.

As described above, the sound for evaluation of various evaluation conditions is based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

(Embodiment 2) FIG. 4 is a block diagram showing a schematic configuration of a sound processing system evaluation apparatus according to a second embodiment of the present invention.

As shown in FIG. 4, the acoustic processing method evaluation apparatus according to the present embodiment includes a data storage means 1 for storing acoustic data and an acoustic data processing for processing a plurality of the acoustic data into evaluation acoustic data. Means 2, an acoustic processing means 3 for processing the acoustic data for evaluation based on an acoustic processing method to be evaluated, and an evaluation output means 4 for outputting an evaluation based on the result processed by the acoustic processing means 3. , And a data input unit 5 for inputting data to be stored in the data storage unit.

The data input section 5 is for inputting data necessary for the evaluation of the acoustic processing method so that the evaluation can be performed under various conditions. For example, the evaluation data can be input by a method of inputting using a storage medium such as a CD-ROM or a method of directly inputting to a buffer of the data storage means 1 using a microphone or the like.

As described above, the sound for evaluation of various evaluation conditions can be calculated based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

(Embodiment 3) FIG. 5 is a block diagram showing a schematic configuration of a sound processing system evaluation apparatus according to a third embodiment of the present invention.

As shown in FIG. 5, the sound processing method evaluation apparatus according to the present embodiment comprises a data storage means 1 for storing sound data, and sound data processing for processing a plurality of the sound data into sound data for evaluation. Means 2, an acoustic processing means 3 for processing the acoustic data for evaluation based on an acoustic processing method to be evaluated, and an evaluation output means 4 for outputting an evaluation based on the result processed by the acoustic processing means 3. A processing content input unit 6 for inputting processing content performed by the acoustic data processing unit.

The processing content input section 6 changes the processing content for calculating the evaluation acoustic data processed by the preset acoustic processing method. By this change, it is possible to change to a processing content that is not set in advance, and it is possible to prepare evaluation data as needed.

As described above, the sound for evaluation of various evaluation conditions is based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

(Embodiment 4) FIG. 6 is a block diagram showing a schematic configuration of a sound processing system evaluation apparatus according to a fourth embodiment of the present invention.

As shown in FIG. 6, the acoustic processing method evaluation apparatus according to the present embodiment includes a data storage means 1 for storing acoustic data and an acoustic data processing for processing a plurality of the acoustic data into acoustic data for evaluation. Means 2, an acoustic processing means 3 for processing the acoustic data for evaluation based on an acoustic processing method to be evaluated, and an evaluation output means 4 for outputting an evaluation based on the result processed by the acoustic processing means 3. A sound processing method input unit 7 for inputting the contents to be processed by the sound processing means from the outside.

The sound processing system loading unit 7 is, for example, a CD-
A program of a sound processing method to be evaluated, such as a voice recognition processing program and a speaker authentication processing program stored in a recording medium such as a ROM, is read.

As described above, the sound for evaluation of various evaluation conditions is based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

(Fifth Embodiment) FIG. 7 is a block diagram showing the schematic arrangement of an acoustic processing method evaluation apparatus according to the fifth embodiment of the present invention.

As shown in FIG. 7, the sound processing method evaluation apparatus according to the present embodiment includes a data storage means 1 for storing sound data, and sound data processing for processing a plurality of the sound data into sound data for evaluation. Means 2, an acoustic processing means 3 for processing the acoustic data for evaluation based on an acoustic processing method to be evaluated, and an evaluation output means 4 for outputting an evaluation based on the result processed by the acoustic processing means 3. , And a weighting means 8 for weighting the acoustic data for evaluation for evaluating the acoustic processing method.

Next, the operation of the weighting process will be described with reference to FIGS.

It is assumed that the sound processing means 3 executes a voice recognition process. Further, the data storage means 1 stores place name data, person name data, and facility name data as evaluation voice data, and an example in which office noise is stored as noise data. In the voice data processing means 2, the in-office noise stored in the data storage means 1 is added to each of the place name data, the person name data, and the facility name data stored in the data storage means 1, and the result is passed to the sound processing means 3. It is like this.

It is assumed that, as a result of executing the voice recognition processing by the acoustic processing means 3, the place name recognition rate is 95%, the person name recognition rate is 90%, and the facility name recognition rate is 98%. Further, as a scale for evaluating the overall speech recognition performance, when the weight of the recognition rate of the place name is 0.5, the weight of the recognition rate of the person name is 0.2, and the weight of the recognition rate of the facility name is 0.3, The evaluation of this voice recognition processing method is 95% × 0.5 + 90% × 0.1.
It can be calculated as 2 + 98% × 0.3 = 94.9%, and this result is output by the evaluation output means 4. In this way, the performance evaluation weighted with respect to the evaluation condition can be performed.

As described above, the sound for evaluation of various evaluation conditions is based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

(Sixth Embodiment) FIG. 9 is a block diagram showing the schematic arrangement of an acoustic processing method evaluation apparatus according to the sixth embodiment of the present invention.

As shown in FIG. 9, the acoustic processing method evaluation apparatus according to the present embodiment includes a data storage means 1 for storing acoustic data and an acoustic data processing for processing a plurality of the acoustic data into evaluation acoustic data. Means 2, an acoustic processing means 3 for processing the acoustic data for evaluation based on an acoustic processing method to be evaluated, and an evaluation output means 4 for outputting an evaluation based on the result processed by the acoustic processing means 3. A weighting unit 8 for weighting the sound processing method and a weight input unit 9 for inputting the weight for each evaluation item are provided.

FIG. 10 is a diagram showing an input example in which weights are input for items displayed on the display. Since the weighting unit 8 is provided with the weight input unit 9, it is possible to specify a weight from the outside and weight the data for evaluation.

As described above, sound for evaluation of various evaluation conditions can be calculated based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

(Embodiment 7) FIG. 11 is a block diagram showing a schematic configuration of a sound processing system evaluation apparatus according to a seventh embodiment of the present invention.

As shown in FIG. 11, the sound processing method evaluation apparatus according to the present embodiment includes a data storage means 1 for storing sound data, and sound data processing for processing a plurality of the sound data into sound data for evaluation. Means 2, an acoustic processing means 3 for processing the acoustic data for evaluation based on an acoustic processing method to be evaluated, and an evaluation output means 4 for outputting an evaluation based on the result processed by the acoustic processing means 3. Is equipped with.

Further, the data storage means 1 has a data input section 5 for inputting the acoustic data to be stored, and the acoustic data processing means 2 has a processing content input section 6 for inputting the processing content. 3 has a sound processing method unit 7 for inputting the contents of the sound processing method, and the weighting means 8 has a weight input unit 9 for inputting a weight for evaluating the sound processing method.

In the present embodiment, the acoustic data of the data storage unit 1, the calculation contents processed by the acoustic data processing means 2, the acoustic processing processed by the acoustic processing means 3, and the weighting weighted by the weighting means 8 are set. Can be specified.

As described above, sound for evaluation of various evaluation conditions can be obtained based on the data stored in the data storage means as a database without preparing a huge amount of sound data for evaluation for evaluating the sound processing method. It is possible to create data, process the created sound data for evaluation by the sound processing method to be evaluated, and evaluate the sound processing method based on the processed result.

[0070]

As described above, according to the present invention, various kinds of sound data for evaluation based on the data stored in the data storage means as a database are prepared without preparing a large amount of sound data for evaluation for evaluating the sound processing method. Acoustic data for evaluation of the evaluation conditions can be created, and the created acoustic data for evaluation can be used to evaluate the sound processing method to be evaluated.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing an acoustic processing method evaluation apparatus according to the first embodiment of the present invention.

FIG. 3 is a block diagram showing an acoustic processing method evaluation apparatus according to the first embodiment of the present invention.

FIG. 4 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a second embodiment of the present invention.

FIG. 5 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a third embodiment of the present invention.

FIG. 6 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a fourth embodiment of the present invention.

FIG. 7 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a fifth embodiment of the present invention.

FIG. 8 is a block diagram showing an acoustic processing method evaluation apparatus according to a fifth embodiment of the present invention.

FIG. 9 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a sixth embodiment of the present invention.

FIG. 10 is a diagram showing a display example of a weight input means of the sound processing method evaluation apparatus according to the sixth embodiment of the present invention.

FIG. 11 is a block diagram showing a schematic configuration of a sound processing method evaluation device according to a seventh embodiment of the present invention.

FIG. 12 is a flowchart showing the operation of the sound processing method evaluation apparatus according to the first embodiment of the present invention.

FIG. 13 is a flowchart showing the operation of the sound processing method evaluation apparatus according to the fifth embodiment of the present invention.

FIG. 14 is a block diagram showing a conventional acoustic processing system evaluation apparatus.

[Explanation of symbols]

1 Data storage means 2 Acoustic data processing means 3 Sound processing means 4 Evaluation output means 5 Data input section 6 Processing content input section 7 Sound processing method input section 8 Weighting means 9 Weight input section 11 voice database 12 Transfer characteristics database 13 noise database 21 Transfer characteristic convolution unit 21a First folding part 21b Second folding part 22 Noise adder 23 Amplitude multiplication section S1 data storage process S2 Acoustic data processing process S3 sound processing process S4 evaluation output process S5 data input process S6 Processing content input process S7 Sound processing system process S8 Weighting process S9 Weight input step S21 Transfer characteristic convolution process S21a First folding step S21b Second folding step S22 Noise addition process S23 Amplitude multiplication step

Claims (24)

[Claims] 1. Data storage means for storing data in advance, acoustic data processing means for processing a plurality of said data into evaluation acoustic data, and said evaluation acoustic data based on an acoustic processing method to be evaluated. An acoustic processing method evaluation device comprising: an acoustic processing unit for processing; and an evaluation output unit for evaluating based on a result processed by the acoustic processing unit and outputting the evaluated result. 2. The data storage means has a data input section for inputting the data.
The acoustic processing method evaluation device described in. 3. The acoustic data processing means has a processing content input section for inputting a processing content for processing the data into acoustic data for evaluation. The acoustic processing method evaluation device described in. 4. The sound processing method according to claim 1, wherein the sound processing means has a sound processing method input unit for inputting a sound processing method for processing the sound data. Processing method evaluation device. 5. Data storage means for pre-storing data, acoustic data processing means for processing a plurality of said data into evaluation acoustic data, and processing contents in which said data is processed into said evaluation acoustic data. Weighting means for weighting the evaluation acoustic data based on the evaluation, acoustic processing means for processing the evaluation acoustic data based on the acoustic processing method to be evaluated, and processed by the acoustic processing means. And a weighting attached to the processed acoustic data for evaluation, and an evaluation output means for outputting the evaluation result. 6. The weighting means according to claim 5, further comprising a weight input section for inputting a weight for evaluation based on a processing content in which the data is processed into the acoustic data for evaluation. Sound processing system evaluation device. 7. The acoustic data processing means includes a transfer characteristic convolution unit for convolving transfer characteristics with voice data stored in the data storage means, and the data storage for the voice data convoluted with the transfer characteristics. 7. The sound processing method evaluation device according to claim 1, further comprising a noise addition unit that adds noise data stored in the means. 8. The acoustic data processing means includes a first transfer characteristic convolution unit that convolves transfer characteristics with voice data stored in the data storage means, and transfer characteristics with noise data stored in the data storage means. A second transfer characteristic convolution unit for convoluting the transfer characteristic, an amplitude multiplication unit for multiplying the convolved noise data by an amplitude, and the noise data in which the transfer characteristic convoluted voice data is multiplied by the amplitude. The sound processing system evaluation device according to claim 1, further comprising: 9. The sound processing unit executes a voice recognition process prepared in advance for the sound data for evaluation, and the evaluation output unit performs the voice recognition based on a result processed by the sound processing unit. 9. The sound processing method evaluation device according to claim 1, wherein the processing is evaluated. 10. The sound processing means executes speaker authentication processing prepared in advance for the sound data for evaluation, and the evaluation output means makes the speech based on a result processed by the sound processing means. 9. The sound processing method evaluation device according to claim 1, wherein the person authentication process is evaluated. 11. The sound processing unit executes a noise suppression process prepared in advance for the sound data for evaluation, and the evaluation output unit performs the noise suppression based on the result processed by the sound processing unit. 9. The sound processing method evaluation device according to claim 1, wherein the processing is evaluated. 12. The sound processing unit executes a sound detection process prepared in advance for the sound data for evaluation, and the evaluation output unit detects the sound based on a result processed by the sound processing unit. 9. The sound processing method evaluation device according to claim 1, wherein the processing is evaluated. 13. A data storage step of pre-storing data in a computer, an acoustic data processing step of processing a plurality of the data into evaluation acoustic data, and an evaluation data processing step based on an audio processing method to be evaluated. An acoustic processing method evaluation program for executing an acoustic processing step of processing acoustic data, and an evaluation output step of evaluating based on a result processed in the acoustic processing step and outputting the evaluated result. 14. The sound processing method evaluation program according to claim 13, which causes a computer to execute a data input step of inputting the data. 15. The method according to claim 13, for causing a computer to execute a processing content input step of inputting a processing content for processing the data into acoustic data for evaluation.
The sound processing method evaluation program according to any one of 1. 16. The sound processing method evaluation program according to claim 13, which causes a computer to execute a sound processing method input step of inputting a processing method for processing the sound data. 17. A data storage step of pre-storing data in a computer, an acoustic data processing step of processing a plurality of the data into acoustic data for evaluation, and the data is processed into acoustic data for evaluation. A weighting step of assigning a weight for evaluation to the evaluation acoustic data based on processing details, an acoustic processing step of processing the evaluation acoustic data based on the acoustic processing method to be evaluated, and the acoustic processing step. An acoustic processing method evaluation program for executing an evaluation based on the result processed in step 1 and the weight attached to the processed audio data for evaluation, and an evaluation output step of outputting the evaluated result. 18. A weight input step of causing a computer to input a weight for evaluation to the acoustic data for evaluation based on a processing content in which the data is processed into the acoustic data for evaluation. 17. A sound processing method evaluation program according to item 17. 19. A transfer characteristic convolution step of convolving a transfer characteristic with the voice data stored in the data storing step, and noise addition for adding noise data to the voice data convoluted with the transfer characteristic in a computer. The sound processing method evaluation program according to any one of claims 13 to 18, which is for executing the steps. 20. A first transfer characteristic convolution step of causing a computer to convolve a transfer characteristic to the voice data stored in the data storing step; and a convolution method of convoluting the transfer characteristic to the noise data stored in the data storing step. 2 transfer characteristic convolution process,
14. A method for performing an amplitude multiplication step of multiplying the convoluted noise data by an amplitude and a noise addition step of adding the convoluted voice data by the amplitude data. Item 19. The sound processing method evaluation program according to any one of Items 18. 21. A computer executes a speech recognition process prepared in advance for the acoustic data for evaluation in the acoustic processing step, and based on a result of the acoustic processing step in the evaluation output step. 21. The sound processing method evaluation program according to claim 13, which evaluates the voice recognition processing. 22. The computer performs a speaker authentication process prepared in advance in the acoustic data for evaluation in the acoustic processing step, and based on a result processed in the acoustic processing step in the evaluation output step. 21. The sound processing method evaluation program according to claim 13, which evaluates the speaker authentication processing. 23. The computer executes a noise suppression process prepared in advance on the acoustic data for evaluation in the acoustic processing step, and based on a result processed in the acoustic processing step in the evaluation output step. 21. The sound processing method evaluation program according to claim 13, which evaluates the noise suppression processing. 24. A computer executes a voice detection process prepared in advance in the acoustic data for evaluation in the acoustic processing step, and based on a result processed in the acoustic processing step in the evaluation output step. 21. The sound processing method evaluation program according to claim 13, which evaluates the voice detection processing.