Audio quality evaluation method and system

Abstract

The invention provides an audio quality evaluation method and system, and belongs to the technical field of audio processing, and the method comprises the steps: extracting the quality feature of a first audio signal, and inputting the quality feature of the first audio signal into a first neural network model to obtain a first quality evaluation score of the first audio signal; converting the first audio signal into a first text, and obtaining a second quality evaluation score of the first audio signal according to the similarity between the first text and an original text corresponding to the first audio signal; and determining a comprehensive quality evaluation score of the first audio signal based on the first quality evaluation score and the second quality evaluation score. According to the audio quality evaluation method and system provided by the invention, manpower and material resources are saved, and the accuracy and reliability of evaluation are improved.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of audio processing, and more specifically, to an audio quality assessment method and system. Background Art

[0002] Audio quality assessment is a key link in the field of audio processing. It is directly related to the fidelity and clarity of audio signals and the user's auditory experience. With the continuous development of audio technology, audio quality assessment is increasingly used in many fields such as radio and television, communications, music production, sound engineering, and daily life. Therefore, how to accurately and efficiently evaluate audio quality has become an important research topic in the field of audio processing technology.

[0003] The existing subjective evaluation methods are resource-consuming, highly subjective, and have poor repeatability; the objective evaluation methods cannot fully reflect the subjective feelings of human hearing. Therefore, an audio quality evaluation method and system are urgently needed to solve the problems in the existing technology. Summary of the invention

[0004] The purpose of the present disclosure is to provide an audio quality assessment method and system to solve the problems in the prior art of wasting resources, being highly subjective, having poor repeatability, and being unable to fully reflect the subjective perception of human hearing.

[0005] According to a first aspect of an embodiment of the present disclosure, there is provided an audio quality assessment method, comprising: Extracting a quality feature of a first audio signal, and inputting the quality feature of the first audio signal into a first neural network model to obtain a first quality assessment score of the first audio signal; Converting the first audio signal into a first text, and obtaining a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal; An integrated quality assessment score of the first audio signal is determined based on the first quality assessment score and the second quality assessment score.

[0006] According to a second aspect of the embodiments of the present disclosure, there is provided an audio quality assessment system, including: a feature extraction module, configured to extract a quality feature of a first audio signal, and input the quality feature of the first audio signal into a first neural network model to obtain a first quality assessment score of the first audio signal; A first quality assessment module, configured to convert the first audio signal into a first text, and obtain a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal; The second quality assessment module is configured to determine a comprehensive quality assessment score of the first audio signal based on the first quality assessment score and the second quality assessment score.

[0007] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the above-mentioned audio quality assessment method when executing the computer program.

[0008] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned audio quality assessment method are implemented.

[0009] The beneficial effects of the audio quality assessment method and system provided by the embodiments of the present disclosure are: On the one hand, the present disclosure achieves a comprehensive and accurate evaluation of the quality of the first audio signal by combining the physical quality characteristics of the first audio signal itself (such as signal-to-noise ratio and distortion) with the accuracy of the content after text conversion (i.e., the similarity with the original text). This multi-dimensional evaluation method effectively avoids the limitations of a single evaluation indicator, ensures that the evaluation results are closer to actual usage, saves manpower and material resources, and improves the accuracy and reliability of the evaluation.

[0010] On the other hand, the present disclosure is not only applicable to basic quality detection of the first audio signal, but also can flexibly adjust the weights of the first and second quality evaluation scores according to the needs of different application scenarios to highlight the evaluation focus of specific aspects. This flexibility enables the method to be more widely applied to various audio processing scenarios to meet the diverse needs of different users. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flowchart of an audio quality assessment method provided by an embodiment of the present disclosure; Figure 2 A structural block diagram of an audio quality assessment system provided by an embodiment of the present disclosure; Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0014] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0015] Please refer to Figure 1 , Figure 1 A flowchart of an audio quality assessment method provided by an embodiment of the present disclosure, the method comprising: S101: extracting quality features of a first audio signal, and inputting the quality features of the first audio signal into a first neural network model to obtain a first quality assessment score of the first audio signal.

[0016] In this embodiment, the first audio signal is audio data acquired by a device in the audio quality assessment system from a certain source, and may be any form of audio content such as recording, music, speech, etc.

[0017] The quality features refer to various attributes or indicators that reflect the quality of the first audio signal, such as signal-to-noise ratio, distortion, clarity, noise level, dynamic range, loudness, spectral characteristics, etc. By extracting these features, the quality of the first audio signal can be quantitatively evaluated, providing basic data for further quality evaluation.

[0018] The first neural network model is a trained artificial neural network model that can be used to process and analyze the quality characteristics of audio signals. The neural network model can establish a complex mapping relationship between audio signal quality characteristics and evaluation scores by learning a large number of audio samples and their corresponding quality evaluation results. When given a new audio signal quality feature, the model can output the corresponding quality evaluation score.

[0019] The first quality evaluation score is the result output by the first neural network model after processing the quality feature of the first audio signal, and is used to quantitatively represent the quality level of the first audio signal. The specific value and range of the evaluation score depend on the training data and design objectives of the neural network model. Generally speaking, a higher evaluation score indicates a better quality of the audio signal, while a lower evaluation score indicates a poorer quality of the audio signal.

[0020] In this embodiment, the system uses signal processing techniques such as Fourier transform and spectrum analysis to extract a series of quality features from the received first audio signal. Then, the system inputs the extracted quality features into a pre-trained first neural network model, and the first neural network model outputs a score, i.e., a first quality assessment score, based on the input quality features. This score can be a numerical value indicating the quality of the audio signal, and a higher numerical value indicates better quality.

[0021] S102: Convert the first audio signal into a first text, and obtain a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal.

[0022] In this embodiment, the first text is a text result obtained by converting the first audio signal through speech recognition technology. The first text is a text representation of the content of the first audio signal. The original text is a known, accurate text content corresponding to the first audio signal, which can be used as a benchmark or reference for evaluating the conversion quality of the first audio signal.

[0023] Similarity is a measure of the similarity between two texts (the first text and the original text). It can be calculated by a variety of algorithms, such as edit distance, cosine similarity, Jaccard similarity, etc.

[0024] The second quality assessment score is a score calculated based on the similarity between the first text and the original text, and is used to quantify the quality of the first audio signal converted into text. A higher score indicates a better conversion quality, that is, the first text is closer to the original text.

[0025] In this embodiment, if the quality of the first audio signal is evaluated only by its physical characteristics (such as signal-to-noise ratio, distortion, frequency response, etc.), it is not enough to ensure the accuracy of the evaluation. Therefore, the first audio signal can be converted into text to evaluate the clarity, recognizability and content accuracy of the audio from a semantic level. The specific working process is as follows: First, the system uses speech recognition technology (such as an automatic speech recognition ASR system) to convert the first audio signal into a first text, and uses the original text corresponding to the first audio signal as a benchmark to evaluate the quality of the converted first text. Secondly, the system uses an appropriate similarity calculation method to compare the first text and the original text, and generates a second quality assessment score based on the similarity calculation result. The second quality assessment score can be combined with the first quality assessment score to jointly judge the quality of the first audio signal.

[0026] S103: Determine a comprehensive quality evaluation score of the first audio signal based on the first quality evaluation score and the second quality evaluation score.

[0027] In this embodiment, in order to conduct a comprehensive and multi-dimensional evaluation of the quality of the audio signal, the system determines the comprehensive quality evaluation score of the first audio signal based on the quality evaluation scores of two different dimensions (the first quality evaluation score and the second quality evaluation score). The first quality evaluation score is usually obtained by analyzing the physical characteristics of the first audio signal (such as signal-to-noise ratio, distortion, frequency response, etc.), which reflects the quality of the first audio signal at the acoustic level. The second quality evaluation score is obtained by converting the first audio signal into text and comparing the similarity with the original text, which reflects the comprehensibility and accuracy of the first audio signal at the semantic level.

[0028] In practical applications, the quality of audio signals is often affected by many factors, including recording environment, equipment performance, transmission conditions, etc. Therefore, a single-dimensional evaluation cannot fully reflect the true quality of the audio signal. This embodiment can more objectively reflect the overall quality level of the audio signal by combining the evaluation scores of different dimensions, providing a basis for subsequent decision-making or optimization.

[0029] There are many ways to determine the comprehensive quality assessment score, such as weighted average method, fuzzy comprehensive evaluation method, neural network fusion method, etc. For example, if the evaluation scores of two dimensions are comparable in importance, a simple arithmetic average or weighted average can be used to calculate the comprehensive score; if the nonlinear relationship or interaction between different dimensions needs to be considered, a more complex fusion method can be used.

[0030] It can be concluded from the above that, on the one hand, the present disclosure achieves a comprehensive and accurate evaluation of the quality of the first audio signal by combining the physical quality characteristics of the first audio signal itself (such as signal-to-noise ratio, distortion) with the accuracy of the content after text conversion (i.e., the similarity with the original text). This multi-dimensional evaluation method effectively avoids the limitations of a single evaluation indicator, ensures that the evaluation results are closer to actual usage, saves manpower and material resources, and improves the accuracy and reliability of the evaluation.

[0031] On the other hand, the present disclosure is not only applicable to basic quality detection of the first audio signal, but also can flexibly adjust the weights of the first and second quality evaluation scores according to the needs of different application scenarios to highlight the evaluation focus of specific aspects. This flexibility enables the method to be more widely applied to various audio processing scenarios to meet the diverse needs of different users.

[0032] In one embodiment of the present disclosure, inputting the quality feature of the first audio signal into the first neural network model to obtain a first quality evaluation score of the first audio signal includes: Inputting each quality feature of the first audio signal into the first neural network model in sequence to obtain a quality assessment score corresponding to each quality feature; The quality assessment score corresponding to each quality feature is weighted to obtain a first quality assessment score.

[0033] In this embodiment, the system extracts multiple quality features from the first audio signal, and these features may include signal-to-noise ratio, spectrum characteristics, distortion, clarity, etc. Each quality feature reflects an aspect of the quality of the first audio signal. Each quality feature of the first audio signal is sequentially input into the first neural network model, and the model independently evaluates each quality feature and outputs a quality evaluation score corresponding to the quality feature.

[0034] Then, the system performs weighted processing on the feature evaluation scores corresponding to each quality feature. The weighted feature evaluation scores are integrated to form a first quality evaluation score of the first audio signal. This score is a single value that integrates the evaluation results of multiple quality features, and can fully reflect the quality of the audio signal at the acoustic level. The purpose of weighting is to adjust the importance of each feature according to the degree of its influence on the overall quality of the first audio signal. This is usually based on prior knowledge, expert experience or data analysis to determine the weight of each feature. It can be concluded from the above that this embodiment can more comprehensively reflect the comprehensive quality of the first audio signal at the acoustic level by individually evaluating each quality feature of the first audio signal and weighted integration, thereby improving the accuracy and meticulousness of the evaluation. At the same time, weighted processing increases the flexibility of the evaluation, and the importance of each feature can be adjusted according to actual needs to meet the quality evaluation needs in different application scenarios.

[0035] In one embodiment of the present disclosure, obtaining a second quality evaluation score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal includes: If the similarity between the first text and the original text corresponding to the first audio signal is less than a first similarity threshold, determining a second quality assessment score of the first audio signal according to the first mapping relationship; If the similarity between the first text and the original text corresponding to the first audio signal is greater than or equal to a first similarity threshold, a second quality assessment score of the first audio signal is determined according to the second mapping relationship.

[0036] In this embodiment, the system may calculate the similarity between the first text and the original text corresponding to the first audio signal by using a text matching algorithm, a natural language processing technique, etc., and then compare the calculated similarity with the first similarity threshold.

[0037] If the similarity is less than the first similarity threshold, it indicates that the degree of match between the first text and the original text is low, and there are large errors or quality problems in the recognition or conversion process of the first audio signal. At this time, the second quality assessment score can be determined according to the first mapping relationship (such as a mapping table or function that tends to be low) to reflect the lower quality level.

[0038] If the similarity is greater than or equal to the first similarity threshold, it indicates that the first text and the original text have a high degree of match, and the first audio signal is relatively accurate in the recognition or conversion process. At this time, a second quality assessment score can be determined according to a second mapping relationship (such as a mapping table or function that tends to be high) to reflect a higher quality level.

[0039] It can be concluded from the above that this embodiment determines the quality assessment score of the first audio signal by segmented determination of the relationship between the similarity and the preset first similarity threshold, and flexibly applies different mapping relationships, thereby effectively improving the accuracy and flexibility of the quality assessment. It can not only quickly identify and feedback quality problems for low similarity situations, but also perform more refined quality assessment for high similarity audio, thereby enhancing the overall performance of the system and user experience.

[0040] In an embodiment of the present disclosure, determining a second quality assessment score of the first audio signal according to the first mapping relationship includes: The first mapping relationship is obtained according to a first formula, and the first formula includes:

[0041] Among them, Q 1 represents the second quality evaluation score of the first audio signal; S represents the similarity between the first text and the original text corresponding to the first audio signal; w represents the amplitude adjustment parameter, which is used to control the amplitude of the quality score change; x represents the drop rate control parameter. When x increases, the quality score decreases faster as the similarity S decreases; y represents the weight parameter of the similarity, which adjusts the degree of influence of the similarity on the quality score. When y increases, a small change in the similarity will have a greater impact on the quality score; z represents the basic score adjustment parameter, which is used to fine-tune the overall quality score.

[0042] A second quality assessment score of the first audio signal is determined according to a first formula.

[0043] In an embodiment of the present disclosure, determining a second quality assessment score of the first audio signal according to the second mapping relationship includes: The second mapping relationship is obtained according to the second formula, and the second formula includes:

[0044] Among them, Q2 represents a second quality evaluation score of the first audio signal, S represents the similarity between the first text and the original text corresponding to the first audio signal, a represents an adjustment coefficient, b represents a sensitivity parameter, and c represents a correction coefficient.

[0045] A second quality assessment score of the first audio signal is determined according to a second formula.

[0046] In this embodiment, when the similarity S increases, the value of (1-S) decreases. Due to the effects of the exponent b and the coefficient c, the value of the entire formula decreases less and less, reflecting that the higher the similarity, the less obvious the change in the quality score, and as the similarity approaches, the quality score approaches 100 points. a can control the magnitude of the decrease in the quality score. When a increases, the magnitude of the decrease in the quality score will increase; when a decreases, the magnitude of the decrease in the quality score will decrease. b determines the sensitivity of the quality score to changes in similarity. When b increases, the quality score is more sensitive to changes in similarity; when b decreases, the quality score is less sensitive to changes in similarity. c is a correction factor, which can be used to fine-tune the quality score to meet different evaluation needs.

[0047] In many practical scenarios, when the audio signal has a low similarity to the original text, it means that there may be a big problem with the accuracy of the audio content. For audio explanations of language learning courses, if the similarity with the textbook text is low, it may make it difficult for learners to accurately understand the content. A lower quality score can remind the producer to make timely improvements to improve the quality of the audio and better meet the needs of learners.

[0048] When the similarity between the audio signal and the original text is already very high, people tend not to feel that the audio quality has been greatly improved or changed. For example, in a news broadcast, if the similarity between the audio and the press release increases from 95% to 98%, the audience may find it difficult to perceive a significant difference in quality. Because in the case of high similarity, the accuracy requirements of information transmission have basically been met, and further slight improvements will not bring about significant perceptual changes. For audio quality assessment systems, when the similarity is at a high level, the quality score does not change significantly, which can make the assessment results more stable.

[0049] For audio quality assessment systems, when the similarity is at a high level, the quality score does not change significantly, which can make the assessment results more stable. This means that even if there are some small errors or fluctuations in the measurement or calculation process, it will not have a big impact on the final quality assessment results. If the quality score changes too significantly at high similarity, it may cause the assessment results to be overly sensitive to some small changes. This may make the assessment system unstable and difficult to accurately reflect the true quality of the audio.

[0050] It can be concluded from the above that, through the mapping relationship and formula of this embodiment, the system can accurately evaluate the quality of the audio signal, especially in terms of content accuracy. Dynamically adjusting the quality score according to the similarity can not only significantly reflect the problem when the similarity is low, but also maintain the stability of the evaluation results when the similarity is high, effectively avoiding oversensitivity caused by small changes. In addition, by adjusting the parameters, it can adapt to different scenarios and needs, improve the applicability and accuracy of the evaluation system, and provide strong support for audio production and evaluation.

[0051] In one embodiment of the present disclosure, the audio quality assessment method further includes: determining a first similarity threshold according to the type of the first audio signal; In response to the first audio signal being of the listening type, determining the first threshold as a first similarity threshold; In response to the first audio signal being of an interactive type, determining the second threshold as a first similarity threshold; The first threshold is smaller than the second threshold.

[0052] In this embodiment, the audio signals of the listening type are mainly those for people to listen to passively, such as clear singing of lyrics in music works, accurate reading of audio books, accurate broadcasting of radio news, etc. In this case, since the user is listening to the audio content attentively, the requirement for the complete correspondence between the audio content and the original text is relatively high.

[0053] For example, when listening to a song, if the singer sings the lyrics inaccurately or makes mistakes, it may affect the audience's experience. For audiobooks, readers can better understand the content of the book by reading the text accurately. In radio news, accurate broadcast content can ensure that the audience gets the correct information.

[0054] Therefore, in the listening type, the setting of the first threshold will be relatively strict. Only when the text converted from the audio signal has a high degree of similarity with the original text, the audio quality will be considered high. This is because in the listening scenario, users pay more attention to the accuracy and completeness of the audio content, and have higher expectations for the correspondence between the audio and the original text. Interactive audio signals usually appear in the interaction between people and devices or between people, such as the response of voice assistants, the voice reply of online customer service, and the dialogue in intelligent voice interactive games.

[0055] In interactive scenarios, the requirement for the audio content to fully correspond to the original text is relatively low. This is because the focus of interaction is often on quickly and effectively conveying information and achieving specific interactive purposes, rather than strict text correspondence. For example, when a user asks a question to a voice assistant, the voice assistant's answer is acceptable as long as it roughly meets the user's needs, even if the expression is not completely consistent with the text of the original question. In the voice response of online customer service, some flexible expressions and explanations are also acceptable as long as they can solve the user's problem.

[0056] In addition, the interaction process usually pays more attention to timeliness and fluency. If the audio content is too consistent with the original text, it may lead to a longer response time and affect the interaction experience. Therefore, under the interaction type, the second threshold is set relatively loosely, allowing a certain degree of text difference to better meet the needs of the interaction scenario.

[0057] In one embodiment of the present disclosure, determining a comprehensive quality assessment score of the first audio signal based on the first quality assessment score and the second quality assessment score includes: Identifying a type of environment in which the first audio signal is located; Determine, according to the environment type, a first weight corresponding to the first quality assessment score and a second weight corresponding to the second quality assessment score; An overall quality assessment score of the first audio signal is determined according to the first quality assessment score, a first weight corresponding to the first quality assessment score, the second quality assessment score, and a second weight corresponding to the second quality assessment score.

[0058] In this embodiment, the system can collect environmental sounds according to the internal sound collection device, determine the location information according to the positioning device, and judge the type of environment according to the environmental sounds and location information. For example, a microphone is used to collect environmental sounds and analyze them, and the noise level, sound frequency distribution and other characteristics are judged to determine whether the environment is noisy; the sensor is used to detect the location information, and the geographic information system is combined to determine whether it is an indoor or outdoor environment.

[0059] Different environments have different perceptions and requirements for audio quality. For example, in a quiet environment, people are more likely to notice subtle flaws in the audio, while in a noisy environment, audio clarity and anti-interference capabilities are more important.

[0060] Once the environment type is determined, the weights of the two quality assessment scores can be determined based on pre-defined rules or models.

[0061] For example, in a quiet indoor environment, more attention is paid to the accuracy of the content of the audio signal. In this case, the second weight corresponding to the second quality evaluation score can be relatively high, while the first weight corresponding to the first quality evaluation score can be appropriately reduced. On the contrary, in a noisy outdoor environment, the technical quality characteristics of the audio, such as clarity and anti-interference ability, are more critical, so the first weight corresponding to the first quality evaluation score can be increased, and the second weight corresponding to the second quality evaluation score can be reduced accordingly.

[0062] The comprehensive quality assessment score can be calculated by the following formula: Comprehensive quality assessment score = first quality assessment score × first weight + second quality assessment score × second weight.

[0063] For example, if the first quality assessment score is 80 points, the corresponding first weight is 0.4; the second quality assessment score is 90 points, the corresponding second weight is 0.6. Then the comprehensive quality assessment score = 80 × 0.4 + 90 × 0.6 = 32 + 54 = 86 points.

[0064] It can be concluded from the above that this embodiment dynamically adjusts the weights of the two quality evaluation scores obtained based on different evaluation dimensions according to the different requirements of different environments for audio quality by identifying the type of environment in which the first audio signal is located, and then obtains a comprehensive quality evaluation score through weighted calculation. This can make the audio quality evaluation more in line with the actual usage scenario and improve the accuracy and practicality of the evaluation.

[0065] In one embodiment of the present disclosure, determining a first weight corresponding to a first quality evaluation score and a second weight corresponding to a second quality evaluation score according to an environment type includes: Determine, according to the environment type, a first initial weight corresponding to the first quality assessment score and a second initial weight corresponding to the second quality assessment score; Determine the noise level of the environment based on the type of environment; The first initial weight is adjusted to a first weight according to the noise value of the environment, and the second initial weight is adjusted to a second weight.

[0066] In this embodiment, the initial weights of the two quality evaluation scores are preliminarily determined according to the type of environment the audio signal is in. The environment type can be divided into different categories such as quiet indoor, noisy indoor, quiet outdoor, and noisy outdoor as analyzed above.

[0067] For different types of environments, a set of initial weight allocation rules can be pre-set. For example, in a quiet indoor environment, the initial weight of the first quality evaluation score (obtained based on the audio signal quality characteristics and the first neural network model) can be preliminarily set to 0.4, and the initial weight of the second quality evaluation score (obtained based on the similarity between the text converted from the audio signal and the original text) can be preliminarily set to 0.6.

[0068] Then, various means are used to determine the noise value of the environment. For example, the ambient sound can be collected through an audio collection device (such as a microphone), and then the intensity, frequency distribution and other characteristics of the sound can be analyzed through a signal processing algorithm to calculate the noise value. If the type of environment is known, the typical noise characteristics of the environment type can also be combined to assist in determining the noise value. For example, in a noisy outdoor environment such as a downtown street, the expected noise value will be relatively high; while in a quiet outdoor environment such as a park, the noise value is relatively low.

[0069] Noise value is an important environmental parameter that directly affects the audibility and quality perception of audio. In a high-noise environment, the clarity and anti-interference ability of audio will be more challenged; while in a low-noise environment, people pay more attention to the details and content accuracy of the audio.

[0070] According to the determined environmental noise value, the previously determined initial weight is adjusted. Generally speaking, the higher the noise value, the greater the weight of the first quality assessment score, while the weight of the second quality assessment score decreases. For example, if the noise value exceeds a certain threshold, the first initial weight can be adjusted from 0.4 to 0.6, and the second initial weight can be adjusted from 0.6 to 0.4. The specific adjustment method can be determined by a preset functional relationship or rule.

[0071] It can be concluded from the above that when determining the comprehensive quality evaluation score of the audio signal, this embodiment first preliminarily determines the initial weights of the two quality evaluation scores based on the environment type, then further determines the noise value of the environment in combination with the environment type, and finally adjusts the initial weight according to the noise value to obtain the final weight.

[0072] The purpose of this is to enable audio quality assessment to fully consider environmental factors, especially the impact of noise on audio quality, so as to more accurately reflect the overall quality of audio in the actual environment. In this way, more targeted quality assessment results can be provided for audio applications in different environments, helping users and developers to better optimize the audio experience and deal with audio-related issues.

[0073] In one embodiment of the present disclosure, the audio quality assessment method further includes: The first weight and the second weight are adjusted according to the user's historical feedback data.

[0074] In this embodiment, the user historical feedback data includes the user's evaluation, rating, opinion, etc. of the audio quality. For example, the user can give a specific satisfaction rating after using the audio service, or put forward opinions on aspects such as audio clarity and content accuracy.

[0075] The system uses data analysis techniques, such as statistical analysis and machine learning algorithms, to analyze these historical feedback data to extract useful information. For example, it analyzes the attention paid to audio technology quality and text similarity in user feedback, as well as the corresponding weight adjustment direction in different feedback situations.

[0076] The system adjusts the first weight and the second weight according to the analysis results. If the user feedback shows that the requirements for audio technology quality are higher, the first weight can be appropriately increased; if the user feedback is more concerned about text similarity, the second weight can be increased.

[0077] User historical feedback data reflects users' actual feelings and needs for audio quality. By adjusting the weights based on these feedback data, audio quality assessment can be made more in line with user expectations and improve user satisfaction. Different users may have different preferences and needs. Considering user feedback can make quality assessment more personalized and adapt to the needs of different user groups.

[0078] It can be concluded from the above that this embodiment introduces user historical feedback data to further optimize the weights, so that the comprehensive quality evaluation score can more accurately reflect the user's actual perception of audio quality. By collecting and analyzing user feedback data and adjusting the weights accordingly, the audio quality evaluation system can be made more intelligent and adaptive, continuously improving the accuracy and practicality of the evaluation, and providing users with a better audio experience.

[0079] In one embodiment of the present disclosure, the audio quality assessment method further includes: The neural network model is trained according to the historical first audio signal and the corresponding quality assessment result to obtain a first neural network model.

[0080] In this embodiment, first, a large number of historical first audio signals are collected. These audio signals can come from different sources, such as music libraries, radio programs, voice call records, etc. At the same time, quality evaluation results corresponding to these historical first audio signals are obtained. The quality evaluation results can be obtained through manual evaluation, existing evaluation methods or other reliable ways.

[0081] Secondly, the collected historical first audio signal and its corresponding quality assessment result are used to train the neural network model. The historical first audio signal is used as input data, and the corresponding quality assessment result is used as the target output. By adjusting the parameters of the neural network, the model can learn the relationship between the characteristics of the audio signal and the quality assessment result. After multiple iterations of training, the performance of the model is continuously optimized until the model achieves good accuracy and generalization ability on the training set. The trained neural network model is the first neural network model, which can be used to perform quality assessment on the new first audio signal.

[0082] It can be concluded from the above that the first neural network model obtained by training the historical first audio signal and the corresponding evaluation result in this embodiment can improve the accuracy and objectivity of audio quality evaluation.

[0083] Corresponding to the audio quality assessment method of the above embodiment, Figure 2 This is a structural block diagram of an audio quality assessment system provided by an embodiment of the present disclosure. For ease of explanation, only the parts related to the embodiment of the present disclosure are shown. Figure 2 The audio quality assessment system 20 includes: a feature extraction module 21, a first quality assessment module 22 and a second quality assessment module 23.

[0084] The feature extraction module 21 is used to extract the quality feature of the first audio signal, and input the quality feature of the first audio signal into the first neural network model to obtain a first quality evaluation score of the first audio signal; A first quality assessment module 22, configured to convert the first audio signal into a first text, and obtain a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal; The second quality evaluation module 23 is configured to determine a comprehensive quality evaluation score of the first audio signal based on the first quality evaluation score and the second quality evaluation score.

[0085] In one embodiment of the present disclosure, the feature extraction module 21 is specifically used for: Inputting each quality feature of the first audio signal into the first neural network model in sequence to obtain a quality assessment score corresponding to each quality feature; The quality assessment score corresponding to each quality feature is weighted to obtain a first quality assessment score.

[0086] In one embodiment of the present disclosure, the first quality assessment module 22 is specifically used to: If the similarity between the first text and the original text corresponding to the first audio signal is less than a first similarity threshold, determining a second quality assessment score of the first audio signal according to the first mapping relationship; If the similarity between the first text and the original text corresponding to the first audio signal is greater than or equal to a first similarity threshold, a second quality assessment score of the first audio signal is determined according to the second mapping relationship.

[0087] In one embodiment of the present disclosure, the second quality assessment module 23 is specifically used to: Identifying a type of environment in which the first audio signal is located; Determine, according to the environment type, a first weight corresponding to the first quality assessment score and a second weight corresponding to the second quality assessment score; An overall quality assessment score of the first audio signal is determined according to the first quality assessment score, a first weight corresponding to the first quality assessment score, the second quality assessment score, and a second weight corresponding to the second quality assessment score.

[0088] In one embodiment of the present disclosure, the second quality assessment module 23 is specifically used to: Determine, according to the environment type, a first initial weight corresponding to the first quality assessment score and a second initial weight corresponding to the second quality assessment score; Determine the noise level of the environment based on the type of environment; The first initial weight is adjusted to a first weight according to the noise value of the environment, and the second initial weight is adjusted to a second weight.

[0089] In one embodiment of the present disclosure, the second quality assessment module 23 is further configured to: The first weight and the second weight are adjusted according to the user's historical feedback data.

[0090] In one embodiment of the present disclosure, the feature extraction module 21 is further used to: The neural network model is trained according to the historical first audio signal and the corresponding quality assessment result to obtain a first neural network model.

[0091] See also Figure 3 , Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. Figure 3 The electronic device 300 in the embodiment shown may include: one or more processors 301, one or more input devices 302, one or more output devices 303 and one or more memories 304. The processors 301, input devices 302, output devices 303 and memories 304 communicate with each other via a communication bus 305. The memory 304 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. The processor 301 is configured to call the program instructions to execute the functions of each module / unit in the above-mentioned system embodiments, such as Figure 2 The functions of modules 21 to 23 are shown.

[0092] It should be understood that in the embodiment of the present disclosure, the processor 301 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0093] The input device 302 may include a touch panel, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and the output device 303 may include a display (LCD, etc.), a speaker, etc.

[0094] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A portion of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.

[0095] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure may execute the implementation methods described in the first and second embodiments of the audio quality assessment method provided in the embodiments of the present disclosure, and may also execute the implementation methods of the electronic device described in the embodiments of the present disclosure, which will not be repeated here.

[0096] In another embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by the processor, all or part of the processes in the above-mentioned embodiment method are implemented, and the computer program can also be completed by instructing the relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0097] The computer-readable storage medium may be an internal storage unit of the electronic device of any of the aforementioned embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0098] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0099] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0100] In the several embodiments provided in the present application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces or units, or it can be an electrical, mechanical or other form of connection.

[0101] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present disclosure.

[0102] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0103] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for evaluating audio quality, characterized in that: include: Extracting a quality feature of a first audio signal, and inputting the quality feature of the first audio signal into a first neural network model to obtain a first quality assessment score of the first audio signal; Converting the first audio signal into a first text, and obtaining a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal; An integrated quality assessment score of the first audio signal is determined based on the first quality assessment score and the second quality assessment score.

2. The audio quality assessment method according to claim 1, wherein: The step of inputting the quality feature of the first audio signal into a first neural network model to obtain a first quality evaluation score of the first audio signal includes: Inputting each quality feature of the first audio signal into the first neural network model in sequence to obtain a quality assessment score corresponding to each quality feature; The quality assessment score corresponding to each quality feature is weighted to obtain the first quality assessment score.

3. The audio quality assessment method according to claim 1, wherein: Obtaining a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal includes: If the similarity between the first text and the original text corresponding to the first audio signal is less than a first similarity threshold, determining a second quality assessment score of the first audio signal according to the first mapping relationship; If the similarity between the first text and the original text corresponding to the first audio signal is greater than or equal to a first similarity threshold, a second quality assessment score of the first audio signal is determined according to a second mapping relationship.

4. The audio quality assessment method according to claim 1, wherein: The determining a comprehensive quality assessment score of the first audio signal based on the first quality assessment score and the second quality assessment score comprises: Identifying the type of environment in which the first audio signal is located; Determine, according to the environment type, a first weight corresponding to the first quality assessment score and a second weight corresponding to the second quality assessment score; A comprehensive quality assessment score of the first audio signal is determined according to the first quality assessment score, a first weight corresponding to the first quality assessment score, the second quality assessment score, and a second weight corresponding to the second quality assessment score.

5. The audio quality assessment method according to claim 4, characterized in that: The determining, according to the environment type, a first weight corresponding to the first quality assessment score and a second weight corresponding to the second quality assessment score includes: Determine, according to the environment type, a first initial weight corresponding to the first quality assessment score and a second initial weight corresponding to the second quality assessment score; Determining a noise value of the environment based on the type of environment; The first initial weight is adjusted to a first weight according to the noise value of the environment, and the second initial weight is adjusted to a second weight.

6. The audio quality assessment method according to claim 4, characterized in that: Also includes: The first weight and the second weight are adjusted according to user historical feedback data.

7. The audio quality assessment method according to claim 1, wherein: Also includes: The neural network model is trained according to the historical first audio signal and the corresponding quality assessment result to obtain a first neural network model.

8. An audio quality assessment system, characterized in that: include: a feature extraction module, configured to extract a quality feature of a first audio signal, and input the quality feature of the first audio signal into a first neural network model to obtain a first quality assessment score of the first audio signal; A first quality assessment module, configured to convert the first audio signal into a first text, and obtain a second quality assessment score of the first audio signal according to a similarity between the first text and an original text corresponding to the first audio signal; The second quality assessment module is configured to determine a comprehensive quality assessment score of the first audio signal based on the first quality assessment score and the second quality assessment score.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images