JP2022181759A - Voice quality evaluation device, voice quality evaluation method, and voice quality evaluation program - Google Patents

Abstract

To provide a voice quality evaluation device, a voice quality evaluation method, and a voice quality evaluation program, which simplify evaluation of transmission quality in voice data transmitted via a network.SOLUTION: A voice quality evaluation device 1 includes: a voice generation unit 12 that converts reference text into reference voice data; a reference voice transmission unit 13 that transmits the reference voice data generated by the voice generation unit to a playback device 50 via a network NW; an evaluation target voice acquisition unit 14 that acquires evaluation target voice data that is played back from the playback device; a voice recognition unit 15 that generates evaluation target text by voice-recognizing words included in the evaluation target voice data; and an evaluation unit 16 that evaluates quality of the voice received via the network based on the evaluation target text.SELECTED DRAWING: Figure 1

Description

The present invention relates to technology for evaluating voice quality.

2. Description of the Related Art In recent years, opportunities to conduct video conferences, remote classes, etc. via communication networks have increased. Therefore, there is a need for a technology that can easily evaluate the quality of this voice.

For example, Patent Document 1 proposes a speech evaluation method for calculating a speech evaluation score by comparing an initial speech recognition result of a sample after error correction and filtering of input speech and an original text after filtering. .
Further, Japanese Patent Application Laid-Open No. 2002-200002 proposes a technique of recording input speech recognition data in a storage unit, matching the speech recognition data with dictionary data by a search unit, and creating a speech recognition result.

JP 2016-51179 A JP 2018-54717 A

A subjective quality evaluation method called MOS (Mean Opinion Score) has long been used to evaluate the voice quality of telephone systems. Also known is an objective quality method POLQA (Perceptual Objective Listening Quality Assessment) that estimates the result of subjective evaluation by MOS using a computer. POLQA is a method of comparing an original voice, called a reference voice, with a voice recorded on the receiver side, and calculating the MOS value.

Here, in voice quality evaluation in video conferences, remote classes, etc., it is important whether the receiving side can accurately hear the contents of the speaker's utterance over several tens of minutes to several hours. However, POLQA is not suitable for evaluating the quality of long-term transmitted speech. POLQA defines standards for reference audio and reference video, and does not cover evaluation of normal conversations and audio that do not meet the standards. In addition, since POLQA is a method of comparing voice data, the amount of data becomes large when trying to evaluate long voice data, and the processing cost is enormous. For this reason, it is difficult to evaluate temporal changes such as, for example, how packetized audio and video are affected by changes in network conditions.

In addition, since POLQA is a method of evaluating the difference between the original speech and the speech on the receiving side, it cannot be said to properly evaluate the accuracy of the utterance content that can be heard by the receiving side. For example, in POLQA, even if processing for improving voice quality such as equalizing processing is performed, a low evaluation is given as a result of determining that there is a difference in voice. Similarly, in situations where spatial noise and sound echoes are included in the collected sound data, such as video conferences and distance learning, even if the noise is removed and played back, there may be differences in the sound. You will be judged and undervalued.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to easily evaluate the quality of transmission of audio data transmitted over a network.

In order to achieve the above object, a speech quality evaluation apparatus according to one aspect of the present invention provides a speech generation unit that converts a reference text into reference speech data; an evaluation target voice acquisition unit that acquires evaluation target voice data played back from the playback device; an evaluation target voice acquisition unit that acquires evaluation target voice data reproduced from the playback device; A speech recognition unit that generates target text, and an evaluation unit that performs quality evaluation of speech received via the network based on the evaluation target text.

The evaluation unit may perform the quality evaluation by comparing the evaluation target text and the reference text.

The speech recognition unit speech-recognizes words included in the reference speech data to generate a second reference text, and the evaluation unit compares the evaluation target text and the second reference text to perform the quality evaluation. may be performed.

The speech recognition unit speech-recognizes words included in the reference speech data to generate a second reference text, and the evaluation unit compares the reference text and the second reference text to perform a first evaluation. A second evaluation may be performed by comparing the reference text and the text to be evaluated, and the quality evaluation may be performed based on the results of the first evaluation and the second evaluation.

In order to achieve the above object, a speech quality evaluation method according to another aspect of the present invention provides speech generation processing for converting a reference text into reference speech data; an evaluation target voice acquisition process for acquiring evaluation target voice data played back from the playback device; and an evaluation target voice data by recognizing words included in the evaluation target voice data. A speech recognition process for generating target text, and an evaluation process for evaluating the quality of speech received via the network based on the evaluation target text.

To achieve the above object, a speech quality evaluation program according to still another aspect of the present invention provides a speech generation instruction for converting a reference text into reference speech data, and the reference speech data generated by the speech generation instruction, a reference voice transmission command to be transmitted to a playback device via a network; an evaluation target voice acquisition command to acquire evaluation target voice data played back from the playback device; A computer is caused to execute speech recognition instructions for generating text to be evaluated and evaluation instructions for evaluating the quality of speech received over the network based on the text to be evaluated.
The computer program can be stored in various data-readable recording media and provided, or can be provided in a downloadable manner via a network such as the Internet.

According to the speech quality evaluation apparatus of the present invention, it is possible to easily evaluate the quality of transmission of speech data transmitted over a network.

1 is a functional block diagram of a speech quality evaluation device according to a first embodiment of the present invention; FIG. An example of data processed in the speech quality evaluation device, including (a) an example of a reference text, (b) an example of reference speech data obtained by reading the reference text, and (c) an evaluation target received via a network. FIG. 4 is a conceptual diagram showing an example of speech data, and (d) an example of text to be evaluated obtained by speech recognition of the speech data received via the network; It is a flowchart which shows the flow of a series of processes performed by the said audio|voice quality evaluation apparatus. 9 is a flow chart showing the flow of a series of processes executed by the speech quality evaluation device according to the second embodiment of the present invention; It is a flowchart which shows the flow of a series of processes performed by the speech quality evaluation apparatus which concerns on the 3rd Embodiment of this invention.

A speech quality evaluation device, a speech quality evaluation method, and a speech quality evaluation program according to embodiments of the present invention will be described below with reference to the drawings.

<First Embodiment>
●Configuration of Voice Quality Evaluation Device As shown in FIG. 1, the voice quality evaluation device 1 is a device that evaluates the quality of voice transmitted and received via the network NW. The network NW may be, for example, the Internet, or an appropriate communication line connected by wire or wirelessly, and may be of any format.

For example, the voice quality evaluation device 1 is connected to the playback device 50 via the network NW. The playback device 50 is, for example, a terminal such as a personal computer, a smart phone, or a tablet, and is a terminal viewed by a viewer of the video call.

The speech quality evaluation device 1 transmits speech data of reference speech (hereinafter also referred to as “reference speech data”) to the reproduction device 50 via the network NW. The reproducing device 50 converts this reference audio data into audio and reproduces it. Note that the reference audio data may be received by the playback device 50 via an appropriate device on the network NW, such as a server. The voice quality evaluation device 1 acquires voice data transmitted via the network NW and reproduced by the reproduction device 50, and evaluates the quality of this voice.

The speech quality evaluation apparatus 1 is configured by a storage medium such as a memory, a processor, a communication module, an input/output interface, etc., and the processor executes a computer program recorded in the storage medium to perform the functions shown in FIG. Blocks are realized. The storage medium is a computer-readable recording medium, and includes storage devices such as RAM (random access memory), ROM (read only memory), disk drives, SSD (solid state drive), and flash memory. may contain. Here, a non-temporary storage device such as a ROM, disk drive, SSD, or flash memory may be included in the voice quality evaluation device 1 as a separate storage device separated from the memory.

The speech quality evaluation apparatus 1 has, for example, mainly a reference text acquisition unit 11, a speech generation unit 12, a reference speech transmission unit 13, an evaluation target speech acquisition unit 14, a speech recognition unit 15, an evaluation unit 16, and the like. Equipped with Part or all of the configuration of the speech quality evaluation apparatus 1 may be realized by another hardware configuration, or part or all of it may be realized by a cloud computer. Also, part of the functions of the voice quality evaluation device 1 may be configured inside the playback device 50 . In this case, for example, the evaluation target speech acquisition unit 14, the speech recognition unit 15, and the evaluation unit 16 may be configured in the playback device 50. FIG.

The reference text acquisition unit 11 is a functional unit that acquires a reference text used for speech quality evaluation. The reference text is, for example, text data as shown in FIG. 2(a), and may be in any language other than Japanese. The reference text acquisition unit 11 may acquire the reference text via an appropriate network, or may receive an input via an appropriate input means of the speech quality evaluation apparatus 1 .

As shown in the conceptual diagram of FIG. 2B, the speech generation unit 12 is a functional unit that converts the reference text acquired by the reference text acquisition unit 11 into reference speech data. The speech generator 12 may read out the reference text with artificial speech and convert it into reference speech data. Further, the speech generation unit 12 may be configured to display the reference text on a display unit such as a display and prompt a speaker, such as an announcer, who speaks accurately, to read the reference text. In this case, the speech generator 12 has a configuration for picking up the speech read out by the speaker.

The reference audio transmission unit 13 is a functional unit that transmits the reference audio data generated by the audio generation unit 12 to the playback device 50 via the network NW. The reproduction device 50 reproduces the received reference audio data as audio. At this time, the playback device 50 may perform appropriate signal processing on the audio data before playback. This signal processing may be, for example, equalizing processing, noise canceling processing, frequency filtering processing, amplification processing, or other acoustic signal processing for making the words contained in the voice more clearly audible. It may be a process of compensating for missing information due to transmission.

Note that the reference audio data may be received by the playback device 50 via an appropriate device on the network NW, such as a server. The signal processing described above may also be applied to the reference audio data by the device.

The evaluation target audio acquisition unit 14 is a functional unit that acquires audio data reproduced from the playback device 50 (hereinafter also referred to as “evaluation target audio data”). The evaluation target speech acquisition unit 14 is connected to the playback device 50 and acquires speech data.
As shown in the conceptual diagram of FIG. 2(c), the evaluation target speech data is partially different from the reference speech data. In the example of the figure, the amplitude of some data shown in area L is reduced. The speech data to be evaluated includes data that has deteriorated more than the reference speech data in the process of transmission, and data that has been processed so that the utterance content can be easily heard by the above-described appropriate signal processing performed in the playback device 50 or devices on the network NW. may be

The speech recognition unit 15 is a functional unit that performs speech recognition of words included in the evaluation target speech data to generate an evaluation target text. In the example of FIG. 2(d), some text shown in area W is different from the reference text.

The evaluation unit 16 is a functional unit that evaluates the quality of speech received via the network NW based on the text to be evaluated. In this embodiment, the evaluation unit 16 compares the evaluation target text and the reference text to evaluate the quality of speech.
A speech quality evaluation score is calculated, for example, by the following equation (1).

Quality evaluation score = Number of sounds correctly recognized in text to be evaluated/Number of sounds in reference text x 100
... (1)

A larger value of the quality evaluation score indicates better speech quality, and the quality evaluation score is 100 when all the sounds in the text to be evaluated are correctly recognized. In the example of FIG. 2, the reference text has 93 sounds, and the text to be evaluated can accurately recognize 89 of them, so the quality evaluation score is 96. The quality evaluation score is displayed on an appropriate display unit that the speech quality evaluation device 1 has or is connected to.

The evaluation unit 16 may calculate one quality evaluation score for the entire evaluation target text, or divide the digestion target text into a plurality of parts, calculate the quality evaluation scores for each of them, and divide them into one reference text. On the other hand, a plurality of quality evaluation scores along the time axis may be calculated. According to the configuration in which a plurality of quality evaluation scores are calculated for one reference text, it is possible to evaluate temporal changes in the transmission state of the network. In this case, the evaluation unit 16 may divide the text to be evaluated so as not to overlap each other, or may divide the text while partially overlapping on the time axis. Note that the speech recognition unit 15 may divide the speech data and then convert each of them into text data. In this case, the evaluation unit 16 evaluates each text data.

●Processing Flow A processing flow for evaluating transmitted voice by the voice quality evaluation apparatus 1 will be described with reference to FIG.
First, the reference text is obtained by the reference text obtaining unit 11 (step S11). Next, the speech generator 12 acquires reference speech data based on the reference text (step S12). Next, the reference voice is transmitted to the playback device 50 by the reference voice transmission unit 13 (step S13).

Next, the target voice data to be played back by the playback device 50 is acquired by the evaluation target voice acquisition unit 14 (step S14). Next, the speech recognition unit 15 performs speech recognition on the target speech data to generate text data (step S15). The evaluation unit 16 evaluates the rate of matching between the reference text and the text to be evaluated, and calculates a quality evaluation score (step S16). Next, the quality evaluation score is displayed on an appropriate display unit (step S17).

According to such a speech quality evaluation apparatus according to the present invention, it is possible to evaluate the quality of transmission over a network, focusing on whether the receiving side can accurately hear what the speaker is saying. Also, according to the speech quality evaluation apparatus of the present invention, since the text data are compared and evaluated, the amount of data to be analyzed can be compressed as compared with the structure that compares the speech data. Therefore, it is possible to evaluate the quality of transmission in long speech. In addition, according to the configuration in which the speech to be evaluated is evaluated as text data, it is possible to appropriately evaluate whether the utterance content is accurately transmitted even when the processing for improving the sound quality is performed.

<Second embodiment>
A voice quality evaluation apparatus according to the second embodiment of the present invention will be described, focusing on the parts different from the first embodiment. In this embodiment, the speech recognition unit 15 speech-recognizes words included in the reference speech data to generate a second reference text, and the evaluation unit 16 compares the evaluation target text with the second reference text to obtain a speech quality evaluation. In addition, description is omitted suitably about the structure similar to 1st Embodiment, and the same code|symbol was attached|subjected.

As shown in FIG. 4, in the speech quality evaluation apparatus according to the second embodiment, after obtaining the reference text (step S11) and generating the reference speech data (step S12), the speech recognition unit 15 generates the reference speech data. is performed to generate a second reference text (step S21). Also, the reference voice data is transmitted (step S13), the evaluation target voice data is acquired via the playback device 50 (step S14), and voice recognition is performed (step S15). The order of step S21 and steps S13 to S15 is arbitrary and may be performed simultaneously.

Next, the evaluation unit 16 evaluates the matching rate between the second reference text and the evaluation target text, calculates the quality evaluation score (step S22), and displays the quality evaluation score on the display unit (step S23).
The quality evaluation score in this case is represented, for example, by Equation (2) below.

Quality evaluation score = Number of sounds correctly recognized in text to be evaluated/Number of sounds in second reference text x 100
... (2)

According to this configuration, when there is an erroneous recognition by the speech recognition unit 15, the same erroneous recognition appears in both the second reference text and the evaluation target text, so the influence of the erroneous recognition on the quality evaluation score can be eliminated. That is, in this embodiment, the voice quality can be evaluated by removing the influence of the voice recognition unit 15 .

<Third Embodiment>
A voice quality evaluation apparatus according to the third embodiment of the present invention will be described, focusing on the parts different from the second embodiment. In this embodiment, the speech recognition unit 15 speech-recognizes words included in the reference speech data to generate a second reference text, and the evaluation unit 16 compares the reference text and the second reference text to obtain a first evaluation. and perform a second evaluation by comparing the reference text and the text to be evaluated, and then evaluate the quality of speech based on the results of the first evaluation and the second evaluation.
It should be noted that the same configurations as those of the first embodiment or the second embodiment will be appropriately omitted from description and given the same reference numerals.

As shown in FIG. 5, in the speech quality evaluation apparatus according to the third embodiment, after obtaining the reference text (step S11) and generating the reference speech data (step S12), the speech recognition unit 15 generates the reference speech data. is performed to generate a second reference text (step S21). Next, a first evaluation is performed to calculate a match rate between the reference text and the second reference text (hereinafter also referred to as "first match rate") (step S31).

Further, the reference speech data is transmitted (step S13), the evaluation target speech data is obtained via the playback device 50 (step S14), and speech recognition is performed (step S15). A second evaluation for calculating a match rate (hereinafter also referred to as a "second match rate") is performed (step S32). The order of steps S21 and S31 and steps S13 to S15 and step S32 is arbitrary and may be performed simultaneously.

Next, the evaluation unit 16 compares the first matching rate and the second matching rate to calculate a quality evaluation score (step S33), and displays this quality evaluation score on the display unit (step S34). For example, the quality evaluation score is represented by Equation (3) below.

Quality evaluation score=second match rate/first match rate×100 (3)

Also, in addition to the quality evaluation score, the first match rate and the second match rate may be displayed on the display unit.

According to this configuration, the accuracy of speech recognition by the speech recognition unit 15 can be confirmed by the first match rate and the second match rate, and the evaluation of voice quality before and after transmission can be checked by the quality evaluation score.

Thus, according to the speech quality evaluation apparatus of the present invention, it is possible to easily evaluate the transmission quality of speech data transmitted via a network.

In this explanation, evaluation of speech quality before and after transmission by a network is described as an example, but the speech quality evaluation device according to the present invention is not limited to networks, and can be used for evaluation of mechanisms in general for transmitting utterance contents. . In addition, the voice quality evaluation device is not only used for network evaluation, but also allows the speaker who is conducting a video conference or a remote class to visually check the part that was not transmitted correctly in real time or after the fact. By constructing a system that prompts the person to repeat the relevant part, it may be possible to help accurate information sharing in a video conference or the like.

1 Speech Quality Evaluation Device 11 Reference Text Acquisition Unit 12 Speech Generation Unit 13 Reference Speech Transmission Unit 14 Evaluation Target Speech Acquisition Unit 15 Speech Recognition Unit 16 Evaluation Unit NW Network

Claims (6)

a speech generator that converts the reference text into reference speech data;
a reference audio transmission unit that transmits the reference audio data generated by the audio generation unit to a playback device via a network;
an evaluation target voice acquisition unit that acquires evaluation target voice data reproduced from the playback device;
a speech recognition unit that performs speech recognition of words included in the evaluation target speech data to generate an evaluation target text;
an evaluation unit that evaluates the quality of speech received via the network based on the text to be evaluated;
comprising
Voice quality evaluation device.
The evaluation unit performs the quality evaluation by comparing the evaluation target text and the reference text.
2. The speech quality evaluation device according to claim 1.
The speech recognition unit speech-recognizes words included in the reference speech data to generate a second reference text,
The evaluation unit performs the quality evaluation by comparing the evaluation target text and the second reference text.
3. The speech quality evaluation device according to claim 1 or 2.
The speech recognition unit speech-recognizes words included in the reference speech data to generate a second reference text,
The evaluation unit compares the reference text with the second reference text to perform a first evaluation, compares the reference text with the evaluation target text to perform a second evaluation, and performs the first evaluation and the second evaluation. Performing the quality evaluation based on the results of the evaluation;
4. The speech quality evaluation device according to any one of claims 1 to 3.
speech generation processing for converting the reference text into reference speech data;
a reference audio transmission process for transmitting the reference audio data generated by the audio generation process to a playback device via a network;
an evaluation target audio acquisition process for acquiring evaluation target audio data reproduced from the playback device;
speech recognition processing for generating an evaluation target text by recognizing words included in the evaluation target speech data;
an evaluation process for evaluating the quality of speech received via the network based on the text to be evaluated;
including,
Speech quality evaluation method.
speech generation instructions for converting the reference text into reference speech data;
a reference audio transmission instruction for transmitting the reference audio data generated by the audio generation instruction to a playback device via a network;
an evaluation target audio acquisition command for acquiring evaluation target audio data reproduced from the playback device;
a speech recognition instruction for generating an evaluation target text by speech recognition of words included in the evaluation target speech data;
an evaluation instruction for evaluating the quality of speech received via the network based on the text to be evaluated;
cause the computer to run
Voice quality evaluation program.

Images