Voice quality inspection method and device and readable storage medium

Abstract

The invention provides a voice quality inspection method and device and a readable storage medium. The method comprises the following steps: acquiring call voice of a quality inspection object; performing timbre feature extraction and content feature extraction on the call voice to obtain voice timbre features and voice content features of the call voice; comparing the voice timbre characteristics of the call voice with standard voice timbre characteristics corresponding to the quality inspection object to obtain a voice timbre characteristic quality inspection result; comparing the voice content features of the call voice with corresponding standard verbal skill content features to obtain a voice content feature quality inspection result; and obtaining a voice quality inspection result of the quality inspection object according to the voice timbre feature quality inspection result and the voice content feature quality inspection result. According to the invention, comprehensive evaluation can be provided for the quality of the voice service, the voice quality inspection efficiency and accuracy are improved, and the improvement of the service quality is facilitated.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a speech quality inspection method, device and readable storage medium. Background Art

[0002] Most companies rely on telephone calls to follow up when handling customer complaints. Customer service and customers mainly communicate through voice calls, and the entire complaint handling process requires quality inspection and supervision. In order to effectively monitor the process and quality of the service, the complaint handling process needs to be recorded in a timely manner. At the same time, the voice quality inspection will conduct a quality inspection on the rationality, process, and content of the solution provided in the complaint recording.

[0003] At present, most quality inspections are conducted through random sampling surveys, manual listening to and processing of recordings, and manual sampling of work order processing procedures to see if the processing content is standardized. However, this method is inefficient and it is difficult to cover all complaint recordings and complaint handling processes.

[0004] In addition, the evaluation criteria for complaint handling recordings vary from person to person. Although manual sampling of recordings can check the necessary actions of the service, there are major flaws in the volume, timbre, and sound quality. Different quality inspectors have different evaluation criteria for the timbre of complaint handlers, and it is difficult to review.

[0005] Therefore, how to improve the efficiency of speech quality inspection and the accuracy of results has become a problem that needs to be solved. Summary of the invention

[0006] The technical problem to be solved by the present application is to provide a speech quality inspection method, device and readable storage medium to solve the problems existing in the prior art in view of the above-mentioned deficiencies in the prior art.

[0007] In a first aspect, the present application provides a method for speech quality inspection, the method comprising:

[0008] S1. Obtain the conversation voice of the inspection object;

[0009] S2, extracting timbre features and content features of the call speech to obtain timbre features and content features of the call speech;

[0010] S3, comparing the voice timbre characteristics of the call voice with the standard voice timbre characteristics corresponding to the quality inspection object to obtain a voice timbre characteristics quality inspection result;

[0011] S4. Compare the voice content features of the call speech with the corresponding standard speech content features to obtain a voice content feature quality inspection result;

[0012] S5. Obtain the speech quality inspection result of the inspection object according to the speech timbre feature quality inspection result and the speech content feature quality inspection result.

[0013] In some embodiments, S2 includes:

[0014] S21, preprocessing the call voice;

[0015] S22, extracting timbre features of the preprocessed call speech to obtain timbre features of the call speech, where the timbre features include time domain features and frequency domain features;

[0016] S23: extract content features of the preprocessed call speech to obtain speech content features of the call speech.

[0017] In some embodiments, in S21, the preprocessing includes at least one of pre-emphasis, framing, windowing, and discrete Fourier transform DFT.

[0018] In some embodiments, in S22, the time domain features include short-time energy, short-time average amplitude, and speech rate of the speech signal;

[0019] The frequency domain features include amplitude, phase, and frequency.

[0020] In some embodiments, S3 includes:

[0021] The voice timbre characteristics of the call speech are compared with the standard voice timbre characteristics corresponding to the quality inspection object by the following formula to obtain the standard deviation of the voice timbre characteristics quality inspection result:

[0022]

[0023] Among them, S 2 is the variance of speech timbre eigenvalue, s n is the voice timbre feature of the call speech, t n is the standard speech timbre feature corresponding to the quality inspection object, and n is the number of feature values.

[0024] In some embodiments, S4 includes:

[0025] The voice content features of the call speech are compared with the corresponding standard speech content features by the following formula to obtain the standard deviation of the voice content feature quality inspection result:

[0026]

[0027] Among them, σ 2 is the variance of speech content feature value, D jis the j-th voice content feature value of the call speech, Cj is the standard speech content feature corresponding to the j-th voice content feature value, and n is the number of feature values.

[0028] In some embodiments, before S1, the method further includes:

[0029] For each quality inspection object, a corresponding standard voice feature archive is established, wherein the standard voice feature archive contains the standard voice timbre features of the quality inspection object;

[0030] A standard speech archive is established, wherein the standard speech archive contains standard speech content features corresponding to different return visit scenarios.

[0031] In a second aspect, the present application provides a speech quality inspection device, the device comprising:

[0032] A voice acquisition module, which is configured to acquire the conversation voice of the quality inspection object;

[0033] A feature extraction module, configured to extract timbre features and content features of the call speech to obtain timbre features and content features of the call speech;

[0034] The speech quality inspection module is configured to compare the speech timbre characteristics of the call speech with the standard speech timbre characteristics corresponding to the quality inspection object to obtain a speech timbre characteristic quality inspection result; compare the speech content characteristics of the call speech with the corresponding standard speech content characteristics to obtain a speech content characteristic quality inspection result; and obtain the speech quality inspection result of the quality inspection object based on the speech timbre characteristic quality inspection result and the speech content characteristic quality inspection result.

[0035] In a third aspect, the present application provides a speech quality inspection device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the speech quality inspection method described in the first aspect above.

[0036] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the speech quality inspection method described in the first aspect is implemented.

[0037] The speech quality inspection method, device and readable storage medium provided by the present application include: obtaining the conversation speech of the quality inspection object; extracting the timbre feature and content feature of the conversation speech to obtain the speech timbre feature and speech content feature of the conversation speech; comparing the speech timbre feature of the conversation speech with the standard speech timbre feature corresponding to the quality inspection object to obtain the speech timbre feature quality inspection result; comparing the speech content feature of the conversation speech with the corresponding standard speech content feature to obtain the speech content feature quality inspection result; according to the speech timbre feature quality inspection result and the speech content feature quality inspection result, the speech quality inspection result of the quality inspection object is obtained. The present application provides a speech quality inspection method, which uses advanced speech analysis technology to compare the speech sample to be tested with the predefined high-quality speech standard when performing the speech quality detection task. This process not only focuses on the consistency of the timbre feature, but also deeply analyzes multiple dimensions of the speech, including but not limited to pitch, volume, speech speed and clarity. Through this method, the difference between each speech sample to be tested and the standard can be quantified, and statistical data reflecting the speech quality, such as variance, can be calculated accordingly. In addition, the method also involves a detailed analysis of the voice content. This includes checking multiple key elements such as the use of question marks, the accuracy of the work number reported, the identification of user questions, the recording of address information, the marking of the time of event occurrence, and the calculation of duration. By comparing the voice content to be tested with the text content of the standard speech, any deviation from the standard can be identified and a detailed quality inspection report can be generated accordingly. Ultimately, the system will provide a comprehensive quality inspection score, which is based not only on the consistency of timbre and content, but also takes into account the fluency of the voice, clarity of expression, and completeness of the information. Such a score can provide a comprehensive assessment of the quality of voice services, improve the efficiency and accuracy of voice quality inspection, and help improve service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0039] Figure 1 A flow chart of a voice quality inspection method provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of the structure of a speech quality inspection device provided in an embodiment of the present application;

[0041] Figure 3 A schematic diagram of the structure of another speech quality inspection device provided in an embodiment of the present application.

[0042] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solution of the present application, the implementation mode of the present application will be further described in detail below with reference to the accompanying drawings.

[0044] It should be understood that the specific embodiments and drawings described herein are only used to explain the present application, rather than to limit the present application.

[0045] It can be understood that, in the absence of conflict, the various embodiments in the present application and the various features in the embodiments can be combined with each other.

[0046] It can be understood that, for the convenience of description, the drawings of the present application only show the parts related to the present application, while the parts unrelated to the present application are not shown in the drawings.

[0047] It can be understood that each unit and module involved in the embodiments of the present application may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0048] It can be understood that the terms "first", "second", etc. in the embodiments of the present application are used to distinguish different objects, or to distinguish different processing of the same object, rather than to describe a specific order of objects.

[0049] It is understandable that, in the absence of conflict, the functions and steps marked in the flowcharts and block diagrams of the present application may occur in an order different from that marked in the drawings.

[0050] It is understood that the flowcharts and block diagrams of the present application illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present application. Among them, each box in the flowchart or block diagram may represent a unit, module, program segment, code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or by a combination of hardware and computer instructions.

[0051] It can be understood that the units and modules involved in the embodiments of the present application can be implemented by software or hardware, for example, the units and modules can be located in a processor.

[0052] When conducting voice quality inspection, it is necessary to comprehensively consider many factors, such as the clarity of the expression content, whether the speaking speed is too fast, whether there is any uncivilized language, as well as subjective, friendly attitude, initiative of service, effectiveness of communication, etc. The above multiple considerations cannot be completed through simple voice-to-text conversion or manual sampling survey.

[0053] At present, the existing voice quality inspection method mainly converts voice into text and simply extracts keywords for quality inspection model matching, and then obtains a comprehensive score. However, the existing method has the following two main disadvantages:

[0054] 1. The most important item in voice quality inspection is ignored in this process: the inspection of timbre. Everyone has different timbre and different voice intensity, so there are defects in meeting the same sound standard requirements.

[0055] 2. The scope of manual voice quality inspection is one-sided. It only conducts random quality inspection on a small number of recordings or fragments in the complaint process through sampling surveys. A large number of long recordings need to be listened to repeatedly by humans, making it impossible to conduct full quality inspection.

[0056] Based on this, the present application proposes a voice quality inspection method, which performs a full-scale quality inspection on telephone recordings in a highly automated and intuitive and visual manner. It can solve the problems of one-sided quality inspection and the human influence on quality inspection results due to individual differences with faster quality inspection and lower economic cost.

[0057] The main concepts of this application include:

[0058] (1) The main technical means used in this method include acoustic testing and spectral analysis, voice outbound call model training and automatic analysis system. The main steps are: input standard voice archive library and preprocessing, extract standard voice feature set and set weights, extract front desk (customer service) recordings and extract front desk voice feature set, compare the front desk voice feature set with the standard voice library, and output voice quality inspection expected value.

[0059] (2) Establish a standard speech text content archive and obtain a standard speech text content feature set, extract the front-end recording and convert it into text content, extract the front-end recording text content feature set, and compare the front-end recording text content with the standard speech text content feature set.

[0060] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0061] The present application provides a speech quality inspection method, the working process of which can be implemented by electronic devices, such as computers, handheld smart terminals, etc. For the convenience of explanation, the embodiments of the present application are described with the method execution subject being a computer.

[0062] In some embodiments, before executing the steps of the speech quality inspection method of the present application, the following pre-steps S01 and S02 are also included:

[0063] S01. For each quality inspection object, a corresponding standard voice feature archive is established, wherein the standard voice feature archive contains the standard voice timbre features of the quality inspection object;

[0064] Specifically, in step S01, a standard voice archive for each voice agent is established for each person being quality inspected, and the voice data is preprocessed, including pre-emphasis, framing, windowing, discrete Fourier transform (DFT) and other operations to improve the quality and readability of the voice signal. The preprocessing steps are as follows:

[0065] S011. Pre-emphasis: In order to emphasize the high frequency part of the standard speech, remove the influence of lip radiation, and increase the high frequency resolution of the speech, pre-emphasis is achieved through the transfer function as a high-pass digital filter.

[0066] H(z)=1-a*z-1

[0067] Where a is the pre-emphasis coefficient, 0.9 <a<1.0。

[0068] Assume that the speech sampling value at time n is x(n), and the result after pre-emphasis processing is y(n)=x(n)-ax(n-1), where a=0.97. The amplitude of the high-frequency part of the spectrum after pre-emphasis is improved.

[0069] S012. Frame division: Since Fourier transform requires the input signal to be stable, but the speech signal is not stable as a whole. If an unstable signal is used as input, Fourier transform will be meaningless. Therefore, it is necessary to process the speech signal by frame division. The frame division of speech signal is actually to perform some transformation or operation on each frame.

[0070] S013, windowing: To emphasize the speech waveform near sample n and weaken the rest of the waveform. Processing each short segment of the speech signal is actually to transform or perform some operation on each short segment. In fact, windowing is equivalent to converting the corresponding element in each frame into the product of the corresponding element of the window sequence. This application uses a Hamming window, and the window function is as follows:

[0071]

[0072] Among them, N is the window length, which is generally equal to the frame length.

[0073] S014. Fast Fourier Transform: Since it is usually difficult to see the characteristics of the signal from the transformation in the time domain, it is usually converted into energy distribution in the frequency domain for observation. Different energy distributions can represent the characteristics of different voices. Therefore, after multiplying the Hamming window, each frame must be subjected to a fast Fourier transform to obtain the energy distribution in the spectrum. The spectrum of each frame is obtained by performing a fast Fourier transform on the framed and windowed signals. Assume that the DFT of the voice signal is:

[0074]

[0075] Wherein, Xn is the input speech signal, and N represents the number of Fourier transform points.

[0076] The power spectrum of the speech signal is obtained by taking the modulus square of the spectrum of the speech signal, and the energy spectrum is calculated using the following formula:

[0077]

[0078] In step S01, after preprocessing is completed, standard voice features of each voice agent are extracted and recorded, including time domain features, frequency domain features, etc., for subsequent analysis and processing.

[0079] Among them, for the time domain feature: the short-time energy En of the n-th frame speech signal xn(m) is:

[0080]

[0081] The short-time average amplitude function Mn is:

[0082]

[0083] In addition, for frequency domain features: fast Fourier transform is performed on the input signal x(n), and the frequency domain features are amplitude, phase, and frequency. The audio is returned by fast Fourier transform as a complex number, where the real part represents the amplitude of the frequency, and the imaginary part represents the phase of the frequency. The standard speech feature values ​​are represented as t1, t2, t3..tn.

[0084] In step S01, the standard voice information archive of the voice agent is in the following format:

[0085] Seat Name XXX Agent Number 40573 …… Short-term energy En …… Short-term average amplitude Mn …… Amplitude A1 …… Phase φ …… frequency f …… Speech rate 120 words / minute

[0086] S02. Establish a standard speech archive, which contains standard speech content features corresponding to different return visit scenarios.

[0087] Specifically, there are different speech standards according to different follow-up scenarios, such as telephone marketing scenarios, bad shutdown follow-up scenarios, network migration, package modification scenarios, and other speech content archives, and standard speech content feature sets are extracted based on various speech standards.

[0088] For example, in the case of mobile network complaint handling, during the telephone follow-up, it is necessary to greet, report the agent's work number, confirm the user's problem phenomenon, problem address, occurrence time, duration, terminal model, complaint network type, etc.; in the mobile network complaint handling scenario, the standard speech content feature set C1 is as follows:

[0089] Scenario Mobile Network …… say hello Hello, XXX …… Reporting number 40573 …… Complaints Slow Internet access …… address Shenzhen City XX …… Time of occurrence 2023 / 11 / 26 / 17:05:07 …… Duration 24 hours …… Terminal Model H203A …… Network Type 5G ……

[0090] In step S02, according to different scenarios, the corresponding scenario speech content feature set and are extracted, which are specifically expressed as follows:

[0091] C1=(x1,x2,x3…xn),1≤n≤T

[0092] Among them, C1 represents the mobile network scenario feature set, n represents the nth feature, and T represents the maximum number of features.

[0093] According to different scenarios, the feature set of standard speech content in the jth scenario is Cj, where n represents the nth feature, as follows:

[0094]

[0095] Figure 1 A schematic diagram of a speech quality inspection method provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the present application provides a speech quality inspection method, which includes S1-S5, as follows:

[0096] S1. Obtain the conversation voice of the inspection object;

[0097] The conversation voice of the quality inspection object may specifically be a recording of a conversation between a voice agent and a customer.

[0098] S2, extracting timbre features and content features of the call speech to obtain timbre features and content features of the call speech;

[0099] In some embodiments, S2 includes:

[0100] S21, preprocessing the call voice;

[0101] In some embodiments, in S21, the preprocessing includes at least one of pre-emphasis, framing, windowing, and discrete Fourier transform DFT.

[0102] In this step, the specific process of pre-processing the call speech of the quality inspection object such as pre-emphasis, framing, windowing, discrete Fourier transform DFT, etc. can refer to the specific processing process in step S01, which will not be repeated here.

[0103] S22, extracting timbre features of the preprocessed call speech to obtain timbre features of the call speech, where the timbre features include time domain features and frequency domain features;

[0104] In some embodiments, in S22, the time domain features include short-time energy, short-time average amplitude, and speech rate of the speech signal;

[0105] The frequency domain features include amplitude, phase, and frequency.

[0106] In this step, the specific process of extracting the timbre features of the pre-processed call speech can refer to the specific processing process in step S01, which will not be repeated here. By extracting the timbre features, the voice timbre features s1, s2, s3...sn of the call speech are obtained.

[0107] S23: extract content features of the preprocessed call speech to obtain speech content features of the call speech.

[0108] After the preprocessing is completed, the call speech is converted into text, and the speech content features of the call speech are extracted according to the processing steps of content feature extraction in step S02, specifically:

[0109]

[0110] Among them, D j represents the sample speech content feature set corresponding to the j scene, and n represents the nth content feature

[0111] S3, comparing the voice timbre characteristics of the call voice with the standard voice timbre characteristics corresponding to the quality inspection object to obtain a voice timbre characteristics quality inspection result;

[0112] In some embodiments, S3 includes:

[0113] The voice timbre characteristics of the call speech are compared with the standard voice timbre characteristics corresponding to the quality inspection object by the following formula to obtain the standard deviation of the voice timbre characteristics quality inspection result:

[0114]

[0115] Among them, S 2 is the variance of speech timbre eigenvalue, s n is the voice timbre feature of the call speech, t nis the standard speech timbre feature corresponding to the quality inspection object, and n is the number of feature values.

[0116] S4. Compare the voice content features of the call speech with the corresponding standard speech content features to obtain a voice content feature quality inspection result;

[0117] In some embodiments, S4 includes:

[0118] The voice content features of the call speech are compared with the corresponding standard speech content features by the following formula to obtain the standard deviation of the voice content feature quality inspection result:

[0119]

[0120] Among them, σ 2 is the variance of speech content feature value, D j is the jth voice content feature value of the call speech, C j is the standard speech content feature corresponding to the j-th speech content feature value, and n is the number of feature values.

[0121] S5. Obtain the speech quality inspection result of the inspection object according to the speech timbre feature quality inspection result and the speech content feature quality inspection result.

[0122] Specifically, the speech quality inspection result of the quality inspection object is obtained according to the ratio of the speech timbre feature quality inspection result and the speech content feature quality inspection result, each accounting for 0.5, and specifically as follows:

[0123] Zn=0.5S 2 +0.5σ 2

[0124] Wherein, Zn is the final speech quality inspection result of the quality inspection object, S 2 is the quality inspection result of speech timbre features, σ 2 is the quality inspection result of the voice content feature.

[0125] This application combines voice timbre feature quality inspection with voice content feature quality inspection to conduct a comprehensive and intuitive quality inspection on all recorded voice seats and effectively analyze these difficult quality inspection issues.

[0126] In some embodiments, it also includes: after obtaining the speech quality inspection results of the quality inspection object (seat), visually displaying the irregularities in the speech and content of each seat, irregular times, irregular scenes, quality inspection scores, rankings, etc. Specifically, it can be divided according to different application scenarios, such as mobile network, telemarketing, etc. It can also be divided according to quality inspection issues such as uncivilized language and too fast speaking speed.

[0127] Through visual display, the test results can be presented in the form of intuitive charts and graphs, which makes it easier for managers to quickly grasp the key indicators of agent performance. At the same time, the rankings and scores generated by the system can be used as the basis for agent training and incentives.

[0128] This application provides a speech quality inspection method, which uses advanced speech analysis technology to compare the speech sample to be tested with the predefined high-quality speech standard when performing the speech quality inspection task. This process not only focuses on the consistency of timbre characteristics, but also deeply analyzes multiple dimensions of speech, including but not limited to pitch, volume, speaking speed and clarity. Through this method, the difference between each speech sample to be tested and the standard can be quantified, and statistical data reflecting speech quality, such as variance, can be calculated based on this.

[0129] In addition, the method also involves a detailed analysis of the speech content. This includes checking the use of question marks, the accuracy of the work number reported, the identification of user questions, the recording of address information, the marking of the time of event occurrence, and the calculation of duration. By comparing the speech content to be tested with the text content of the standard speech, any deviation from the standard can be identified and a detailed quality inspection report can be generated accordingly.

[0130] Finally, the system will provide a comprehensive quality inspection score, which is not only based on the consistency of timbre and content, but also takes into account the fluency of speech, clarity of expression and completeness of information. Such a score can provide a comprehensive evaluation of the quality of voice services, improve the efficiency and accuracy of voice quality inspection, and help improve service quality.

[0131] The advantages of the solution proposed in this application are:

[0132] 1. It can conduct quality inspection on all recorded voices, thus avoiding defects caused by sampling a small amount of voice or voice fragments.

[0133] 2. This application uses each individual's own timbre, volume, speaking speed and other characteristics as the standard recording library, and uses this for comparison to avoid unreasonable quality inspection due to individual voice differences and reduce human factors.

[0134] 3. Using the standard script text content as the text content archive, convert the front-end voice into text for comparison. In this way, the quality inspection results of voice features and content features can be expected, and the full-scale automated quality inspection can be achieved and the quality inspection results can be visualized.

[0135] It should be understood that, although the various steps in the flowcharts in the above-described embodiments are sequentially displayed according to the indications of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and they can be executed in other orders. Moreover, at least a portion of the steps in the figure may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily to be carried out sequentially, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0136] Figure 2 A schematic diagram of a speech quality inspection device provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the present application provides a speech quality inspection device, the device comprising:

[0137] A voice acquisition module 11, which is configured to acquire the conversation voice of the quality inspection object;

[0138] A feature extraction module 12, which is configured to extract timbre features and content features of the call speech to obtain timbre features and content features of the call speech;

[0139] The speech quality inspection module 13 is configured to compare the speech timbre characteristics of the call speech with the standard speech timbre characteristics corresponding to the quality inspection object to obtain a speech timbre characteristic quality inspection result; compare the speech content characteristics of the call speech with the corresponding standard speech content characteristics to obtain a speech content characteristic quality inspection result; and obtain the speech quality inspection result of the quality inspection object based on the speech timbre characteristic quality inspection result and the speech content characteristic quality inspection result.

[0140] Regarding the limitation of the speech quality inspection device, reference may be made to the limitation of the speech quality inspection method in the above-mentioned embodiments of the present application, which will not be elaborated in this embodiment.

[0141] Figure 3 Another schematic diagram of the speech quality inspection device provided in the embodiment of the present application is as follows Figure 3 As shown, in some embodiments, the present application provides a speech quality inspection device, including a memory 22 and a processor 21, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the speech quality inspection method in the above-mentioned embodiments of the present application.

[0142] The memory is connected to the processor, the memory may be a flash memory or a read-only memory or other memory, and the processor may be a central processing unit or a single-chip microcomputer.

[0143] In some embodiments, the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the speech quality inspection method in the above-mentioned embodiments of the present application is implemented.

[0144] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0145] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present application, but the present application is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of the present application, and these modifications and improvements are also considered to be within the scope of protection of the present application.

Claims

1. A speech quality inspection method, characterized in that: The method comprises: S1. Obtain the conversation voice of the inspection object; S2, extracting timbre features and content features of the call speech to obtain timbre features and content features of the call speech; S3, comparing the voice timbre characteristics of the call voice with the standard voice timbre characteristics corresponding to the quality inspection object to obtain a voice timbre characteristics quality inspection result; S4. Compare the voice content features of the call speech with the corresponding standard speech content features to obtain a voice content feature quality inspection result; S5. Obtain the speech quality inspection result of the inspection object according to the speech timbre feature quality inspection result and the speech content feature quality inspection result.

2. The method for voice quality inspection according to claim 1, characterized in that: S2, including: S21, preprocessing the call voice; S22, extracting timbre features of the preprocessed call speech to obtain timbre features of the call speech, where the timbre features include time domain features and frequency domain features; S23: extract content features of the preprocessed call speech to obtain speech content features of the call speech.

3. The method for voice quality inspection according to claim 2, characterized in that: In S21, the preprocessing includes at least one of pre-emphasis, frame division, windowing, and discrete Fourier transform DFT.

4. The method for voice quality inspection according to claim 2, characterized in that: In S22, the time domain features include short-time energy, short-time average amplitude, and speech rate of the speech signal; The frequency domain features include amplitude, phase, and frequency.

5. The method for voice quality inspection according to claim 1, characterized in that: S3, including: The voice timbre characteristics of the call speech are compared with the standard voice timbre characteristics corresponding to the quality inspection object by the following formula to obtain the standard deviation of the voice timbre characteristics quality inspection result: Among them, S 2 is the variance of speech timbre eigenvalue, s n is the voice timbre feature of the call speech, t n is the standard speech timbre feature corresponding to the quality inspection object, and n is the number of feature values.

6. The method for voice quality inspection according to claim 1, characterized in that: S4, including: The voice content features of the call speech are compared with the corresponding standard speech content features by the following formula to obtain the standard deviation of the voice content feature quality inspection result: Among them, σ 2 is the variance of speech content feature value, D j is the j-th voice content feature value of the call speech, Cj is the standard speech content feature corresponding to the j-th voice content feature value, and n is the number of feature values.

7. The method for speech quality inspection according to any one of claims 1 to 6, characterized in that: Before S1, it also included: For each quality inspection object, a corresponding standard voice feature archive is established, wherein the standard voice feature archive contains the standard voice timbre features of the quality inspection object; A standard speech archive is established, wherein the standard speech archive contains standard speech content features corresponding to different return visit scenarios.

8. A speech quality inspection device, characterized in that: The device comprises: A voice acquisition module, which is configured to acquire the conversation voice of the quality inspection object; A feature extraction module, configured to extract timbre features and content features of the call speech to obtain timbre features and content features of the call speech; The speech quality inspection module is configured to compare the speech timbre characteristics of the call speech with the standard speech timbre characteristics corresponding to the quality inspection object to obtain a speech timbre characteristic quality inspection result; compare the speech content characteristics of the call speech with the corresponding standard speech content characteristics to obtain a speech content characteristic quality inspection result; and obtain the speech quality inspection result of the quality inspection object based on the speech timbre characteristic quality inspection result and the speech content characteristic quality inspection result.

9. A speech quality inspection device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the speech quality inspection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the speech quality inspection method according to any one of claims 1 to 7 is implemented.

Images