JP2002064612A - Voice sample gathering method for subjective quality estimation and equipment for executing the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice sample gathering method for subjective quality estimation which gathers a suitable number of voice samples, and equipment for executing the method. SOLUTION: This equipment for executing a voice sample gathering method is provided with an original voice input device 1; a communication voice input device 2 which inputs a communication voice obtained by reproducing a communication signal via a transmission line, and generates a communication voice signal as voice samples; a deterioration time length calculating device 3 which inputs an original voice signal and the voice samples, compares them, and calculates a rate of time length where the compared result exceeds a prescribed range to the total time length of the inputted original voice, as a deterioration time length; a judgment coefficient calculating device 4 for calculating a judgment coefficient from a set of deterioration time lengths; and a gathering work end judging device 5 for performing judgment that the gathering work is ended when the judgment coefficient is in a range of a prescribed value set by a work practitioner, and the gathering work is continued when the coefficient is outside the range.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for recording a speech sample for subjective quality evaluation and an apparatus for implementing the method, and more particularly, to a subjective quality evaluation used for estimating the speech quality of a voice communication service by subjective quality evaluation. The present invention relates to an audio sample recording method and a device for executing the method.

[0002]

2. Description of the Related Art To estimate the speech quality of a voice communication service by subjective quality evaluation, an appropriate number of voice samples are required as a subject of a subjective quality evaluation test. Judgment was decided based on subjectivity.

[0003]

As an example, in a voice communication service using a mobile phone or an Internet phone, there is a variation in the occurrence of deterioration in speech quality. That is, in a voice communication service using a mobile phone or an Internet phone, the call status changes drastically with time.
If you ’re trying to estimate the call quality of your voice service,
Since only a limited number of audio samples are recorded and used for the subjective evaluation test, only the samples that happen to contain a lot of deterioration are used as evaluation targets, and conversely, there is almost no deterioration It is assumed that only samples are used as evaluation targets. Therefore, there may be a case where the substance of the speech quality of the voice communication service cannot be properly grasped.

[0004] By recording a sufficient number of voice samples and performing a subjective quality evaluation, it is possible to guarantee an appropriate grasp of the actual communication quality of the voice communication service.
However, since the occurrence rate and variance of voice communication service degradation are not known, it is necessary to make a general judgment as to how many voice samples should be prepared and sufficient for proper subjective quality evaluation. Can not. From the viewpoint of effective use of resources required for audio recording and subjective quality evaluation experiments, it is desirable that the number of audio samples to be recorded is a necessary and sufficient number.

According to the present invention, an original voice input to a voice communication service is compared with a voice sample which is a call signal recorded after passing through the voice communication service, and the obtained comparison result satisfies a predetermined condition. In this case, it is determined that the number of audio samples recorded so far is a necessary and sufficient number, and by stopping further recording of the audio samples, a voice sample recording method for subjective quality evaluation that has solved the above-described problem,
And an apparatus for implementing the same.

[0006]

Means for Solving the Problems Claim 1: An original voice input to a voice communication service is compared with a voice sample which is a call signal recorded after passing through the voice communication service, and a comparison result obtained is obtained. When a predetermined condition is satisfied, it is determined that the number of voice samples recorded so far is a necessary and sufficient number, and a voice sample recording method for stopping further recording of voice samples is configured.

In a preferred embodiment of the present invention, a portion where the comparison result between the original voice and the voice sample exceeds a specified value range is regarded as a degraded portion, and one voice sample is recorded. The ratio of the degraded part in the time length is stored for the recorded audio sample, and the judgment coefficient is obtained based on the stored ratio of the degraded part, and the judgment coefficient falls within the range of the specified value set by the worker. Is set as a predetermined condition for judging that the number of recorded audio samples is necessary and sufficient.

Further, in the voice sample recording method according to the present invention, the time length of one voice sample is divided into a plurality of sections of a fixed time, and the S / N ratio is calculated for each divided section. The sample recording method was configured. here,
Claim 4: An original audio input device 1 for inputting an original audio and generating an original audio signal corresponding thereto, and inputting a call audio obtained by reproducing an incoming call signal transmitted through a transmission path and inputting a call audio. A speech input device 2 for generating a signal as a speech sample is provided. An original speech signal obtained from the original speech input device 1 and a speech sample obtained from the speech input device 2 are input and compared. And a deterioration time length calculating device 3 for calculating a ratio of the time length exceeding the set range and the total time length of the input original sound as a deterioration time length. The deterioration time length obtained by the deterioration time length Remember,
A judgment coefficient calculation device 4 for calculating a judgment coefficient from a set of stored deterioration time lengths, wherein the judgment coefficient obtained by the judgment coefficient calculation device 4 is within a range of a prescribed value set by an operator; Constituted a voice sample recording device provided with a recording work end determination device 5 for ending the recording work and determining to continue the recording work if it is out of the range.

[0010] Claim 5: In the audio sample recording apparatus according to claim 4, the deterioration time length calculating device 3
Constituted an audio sample recording apparatus having a configuration in which the time length of one audio sample was divided into a plurality of sections of a fixed time, and the S / N ratio was calculated for each divided section.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the embodiment shown in FIG. In FIG. 1, reference numeral 1 denotes an original voice input device, 2 denotes a call voice input device, 3 denotes a deterioration time length calculating device, 4 denotes a judgment coefficient calculating device, and 5 denotes a recording work end determining device. The original sound input device 1 is an acousto-electric conversion device that inputs an original sound and generates an original sound signal corresponding thereto. The original audio signal is a signal before transmission on the transmission path,
It does not include transmission distortion and noise. An original audio is input to the original audio input device 1, and an original audio signal corresponding to the original audio is generated.

The speech input device 2 receives a speech reproduced from a speech signal containing transmission distortion and noise arriving through a transmission line, and generates a speech signal corresponding to the speech as a speech sample. It is an electric conversion device. The degradation time length calculation device 3 receives the original voice signal obtained from the original voice input device 1 and the speech voice signal obtained from the speech voice input device 2 and compares them, and the value of the comparison exceeds the specified range. This is a device that calculates the ratio of the total time length of the input original audio to the total time length of the input original sound as the deterioration time length.

The judgment coefficient calculating device 4 is a device that stores the deterioration time length obtained by the deterioration time length calculating device 3 and calculates a judgment coefficient from a set of the stored deterioration time lengths. The recording operation end determination device 5 ends the recording operation when the determination coefficient obtained by the determination coefficient calculation device 4 is within the range of the specified value set by the operation executor. It is a device that determines to continue. here,
The operation of the embodiment will be described.

First, an original audio prepared for recording an audio sample is input to an original audio input device 1, and a reproduced audio obtained through an audio communication service to be evaluated using the obtained original audio signal. Is input to the call voice input device 2. The original audio signal obtained by the original audio input device 1 and the reproduced audio signal obtained by the call audio input device 2 are respectively sampled at a specific sampling frequency and converted into digital data.

The two digital data are used as the deterioration time length calculating device 3
And the S / N ratio is calculated. The calculation of the S / N ratio is performed for each divided section by dividing the duration of one audio sample into a plurality of sections of a fixed time of about 0.1 sec. The S / N ratio obtained for each divided section is compared with a specified value, and if the S / N ratio exceeds the specified value, it is determined as a deteriorated portion. The determination of the deteriorated portion is performed over the entire time length of one audio sample, and the ratio of the time length of the entire deteriorated portion to the total time length of the original voice is calculated.

The deteriorated portion calculated by the deterioration time length calculation device 3 is input to the judgment coefficient calculation device 4 and stored therein. Every time a deteriorated portion is input from the deterioration time length calculating device 3, an average value and a confidence interval of the deteriorated portion are obtained together with the already stored deteriorated portion. A determination coefficient (= confidence interval / average value of deteriorated portion) is determined based on the determined average value of the deteriorated portion and the confidence interval. Here, the confidence interval means that the measured degradation time length falls within the range with certain certainty. And the judgment coefficient means the magnitude of the fluctuation obtained from each measured deterioration time length.

The judgment coefficient obtained by the judgment coefficient calculating device 4 is inputted to the recording work end judging device 5, and if the judgment coefficient is within the specified value determined by the recording work performer, this is the transmission to be measured. Since it means that the audio sample enough to cover the degradation pattern that can be measured from the road is obtained, the end of the recording operation is displayed, and if it is outside the predetermined value, the continuation of the recording operation is displayed. As described above, it is determined whether or not a sufficient number of voice samples necessary for accurately grasping the actual communication quality of the voice communication service to be evaluated have been obtained.

[0017]

As described above, according to the present invention, the original voice input to the voice communication service is compared with the voice sample, which is the call signal recorded after passing through the voice communication service, and is obtained. If the obtained comparison result satisfies a predetermined condition, it is determined that the number of audio samples recorded so far is a necessary and sufficient number, and by adopting a configuration in which further recording of audio samples is terminated, the audio sample is discontinued. It is possible to easily record a sufficient number of voice samples necessary for accurately grasping the actual communication quality of the communication service. In other words, the task performer of the speech quality grasping work of the voice communication service can perform the recording and the evaluation test of an appropriate scale, thereby minimizing the cost of these works. .

[Brief description of the drawings]

FIG. 1 illustrates an embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Original voice input device 2 Call voice input device 3 Degradation time length calculation device 4 Judgment coefficient calculation device 5 Recording end judgment device

Claims (5)

[Claims] 1. A method for comparing an original voice input to a voice communication service with a voice sample which is a call signal recorded after passing through the voice communication service, and when the obtained comparison result satisfies a predetermined condition, A sound sample recording method characterized by judging that the number of sound samples recorded so far is a necessary and sufficient number and terminating further recording of the sound samples. 2. The audio sample recording method according to claim 1, wherein a portion where the comparison result between the original audio and the audio sample exceeds a specified value range is regarded as a deteriorated portion and occupies a time length of one audio sample. The ratio of the degraded part is stored for the recorded voice sample, and a judgment coefficient is obtained based on the stored ratio of the degraded part, and it is recorded that the judgment coefficient is within the range of the specified value set by the worker. A sound sample recording method, characterized in that a predetermined condition for judging that the number of sound samples obtained is necessary and sufficient. 3. The audio sample recording method according to claim 2, wherein a time length of one audio sample is divided into a plurality of sections of a fixed time, and an S / N ratio is calculated for each divided section. How to record audio samples. 4. An original voice input device for inputting an original voice and generating an original audio signal corresponding thereto, and inputting a call voice obtained by reproducing an incoming call signal transmitted through a transmission path and inputting a call voice A voice input device for generating a signal as a voice sample; a voice signal obtained from the original voice input device and a voice sample obtained from the voice input device are input and compared;
A degradation time length calculation device that calculates a ratio of a time length in which the comparison result exceeds a specified range and a total time length of the input original sound as a degradation time length, and a degradation time obtained by the degradation time length calculation device. A determination coefficient calculating device for storing the length and calculating a determination coefficient from a set of the stored deterioration time lengths, wherein the determination coefficient obtained by the determination coefficient calculation device is within a range of a specified value set by a work executor. A recording operation end determining device for ending the recording operation when the condition is not satisfied and continuing the recording operation when out of the range. 5. The audio sample recording device according to claim 4, wherein the degradation time length calculation device divides a time length of one audio sample into a plurality of sections of a fixed time, and sets an S / N ratio for each division section. An audio sample recording device having a configuration for calculating.