CN111326169A - A method and device for evaluating voice quality - Google Patents

Abstract

The invention discloses a voice quality evaluation method and device. The voice signal to be evaluated is obtained, the voice signal to be evaluated is compared with the stored voice signal, and when the difference between the two is relatively large, the built-in voice quality is evaluated. The model is updated to obtain a new speech quality evaluation model. The new speech quality evaluation model is used to evaluate the speech signal to be evaluated. By continuously learning the speech signal, the speech quality evaluation model is continuously updated, thereby improving the accuracy of speech evaluation.

Description

A method and device for evaluating voice quality

technical field

The present invention relates to the technical field of communications, and in particular, to a method and device for evaluating voice quality.

Background technique

Internet-based voice service has become one of the important services of the network, and it is the focus of various service providers. Voice quality is an important factor in evaluating the quality of communication networks. In order to achieve the purpose of evaluating voice quality, develop effective voice quality Evaluation methods are indispensable.

At present, the speech evaluation method usually uses a fixed speech quality evaluation model to evaluate the speech quality. The specific method is as follows: extract the characteristic parameters of the speech signal, and train the speech quality evaluation model based on the extracted characteristic parameters, and use the speech quality evaluation model. Evaluate the speech signal, because the speech quality evaluation model trained in this way is suitable for scenarios where the speech environment does not change much, and the model is fixed and cannot be changed. This method of evaluating the speech quality evaluation model may make the accuracy of speech evaluation relatively low.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a speech quality evaluation method and device to improve the accuracy of speech evaluation.

The purpose of this invention is to realize through the following technical solutions:

In a first aspect, the present invention provides a method for evaluating voice quality, including:

Obtain the voice signal to be evaluated, and determine the identification information of the voice signal to be evaluated;

If it is determined that the identification information of the voice signal to be evaluated is different from the identification information of the stored voice signal, the voice signal to be evaluated is regarded as a new voice signal, and when the number of the new voice signal is greater than that of the first voice signal When a preset threshold is used, the first voice quality evaluation model is updated to obtain a second voice quality evaluation model;

Wherein, the stored voice signal is the voice signal obtained before the voice signal to be evaluated;

The speech signal to be evaluated is evaluated by using the second speech quality evaluation model.

Optionally, the first voice quality evaluation model is updated to obtain a second voice quality evaluation model, including:

obtaining the characteristic parameters of the new speech signal;

Use decision tree algorithm to train the feature parameters, update the first voice quality evaluation model, and obtain a second voice quality evaluation model;

The characteristic parameters include at least one of the following: signal-to-noise ratio, background noise, noise level, asymmetric interference value of the average speech signal spectrum, high-frequency flatness analysis, spectrum level range, spectrum level standard deviation, relative noise floor, linearity Skewness coefficient of prediction coefficients, cepstral skewness coefficient, voiced sound, mean cross section of back cavity, vocal tract amplitude variation, speech level.

Optionally, after acquiring the speech signal to be evaluated, the method further includes:

At least one of the following preprocessing is performed on the voice signal to be evaluated: voice data validity detection, voice data normalization processing, and default value difference fitting.

Optionally, after acquiring the speech signal to be evaluated, the method further includes:

Evaluate the to-be-evaluated speech signal according to the first speech quality evaluation model to obtain the speech quality of the to-be-evaluated speech signal;

The speech quality of the speech signal to be evaluated is classified to obtain speech quality of different interval levels; the speech quality of different interval levels is used to characterize the speech quality of different categories.

Optionally, the identification information of the voice signal to be evaluated is different from the identification information of the saved voice signal, including:

The speech quality of the speech signal to be evaluated is different from the speech quality of the already stored speech signal, and/or the characteristic parameter of the speech signal to be evaluated is different from the characteristic parameter of the already stored speech signal.

Optionally, the voice quality of the voice signal to be evaluated is different from the voice quality of the voice signal that has been saved, including:

The voice quality of the voice signal to be evaluated and the voice quality of the voice signal that has been saved are voice signals of the same interval level, and the voice quality of the voice signal to be evaluated is the same as the voice signal of the voice signal that has been saved. The difference in quality is less than the second preset threshold; or

The voice quality of the voice signal to be evaluated and the voice quality of the saved voice signal are voice signals of different interval levels.

In a second aspect, the present invention provides a device for evaluating voice quality, comprising:

an acquisition unit for acquiring the speech signal to be evaluated;

The determining unit is used to determine the identification information of the voice signal to be evaluated, and when it is determined that the voice quality of the voice signal to be evaluated is different from the voice quality of the voice signal that has been saved, the voice signal to be evaluated is as a new voice signal;

an updating unit, when the number of the new voice signals is greater than the first preset threshold, updating the first voice quality evaluation model to obtain a second voice quality evaluation model;

Wherein, the stored voice signal is the voice signal obtained before the voice signal to be evaluated;

An evaluation unit, configured to evaluate the speech signal to be evaluated by using the second speech quality evaluation model.

Optionally, the updating unit is specifically configured to update the first voice quality evaluation model as follows to obtain a second voice quality evaluation model:

obtaining the characteristic parameters of the new speech signal;

Use decision tree algorithm to train the feature parameters, update the first voice quality evaluation model, and obtain a second voice quality evaluation model;

The characteristic parameters include at least one of the following: signal-to-noise ratio, background noise, noise level, asymmetric interference value of the average speech signal spectrum, high-frequency flatness analysis, spectrum level range, spectrum level standard deviation, relative noise floor, linearity Skewness coefficient of prediction coefficients, cepstral skewness coefficient, voiced sound, mean cross section of back cavity, vocal tract amplitude variation, speech level.

Optionally, the apparatus further includes a processing unit for:

At least one of the following preprocessing is performed on the voice signal to be evaluated: voice data validity detection, voice data normalization processing, and default value difference fitting.

Optionally, the evaluation unit is also used for:

Evaluate the to-be-evaluated speech signal according to the first speech quality evaluation model to obtain the speech quality of the to-be-evaluated speech signal;

The processing unit is also used to:

The voice quality of the voice signal to be evaluated is classified to obtain the voice quality of different interval levels; the voice quality of different interval levels is used to represent the voice quality of different categories.

Optionally, the identification information of the voice signal to be evaluated is different from the identification information of the saved voice signal, including:

The speech quality of the speech signal to be evaluated is different from the speech quality of the already stored speech signal, and/or the characteristic parameter of the speech signal to be evaluated is different from the characteristic parameter of the already stored speech signal.

Optionally, the voice quality of the voice signal to be evaluated is different from the voice quality of the voice signal that has been saved, including:

The voice quality of the voice signal to be evaluated and the voice quality of the voice signal that has been saved are voice signals of the same interval level, and the voice quality of the voice signal to be evaluated is the same as the voice signal of the voice signal that has been saved. The difference in quality is less than the second preset threshold; or

The voice quality of the voice signal to be evaluated and the voice quality of the saved voice signal are voice signals of different interval levels.

In a third aspect, the present invention also provides a device for evaluating voice quality, including:

memory for storing program instructions;

The processor is configured to call the program instructions stored in the memory, and execute the method of the first aspect according to the obtained program.

In a fourth aspect, the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores computer instructions, which, when executed on a computer, cause the computer to execute the method described in the first aspect.

The present invention provides a voice quality evaluation method and device. The voice signal to be evaluated is obtained, the voice signal to be evaluated is compared with the stored voice signal, and when the difference between the two is relatively large, the built-in voice quality evaluation model is evaluated. Update, get a new voice quality evaluation model, use the new voice quality evaluation model to evaluate the voice signal to be evaluated, and continuously update the voice quality evaluation model by continuously learning the voice signal, thereby improving the accuracy of the voice evaluation.

Description of drawings

1 is a flowchart of a method for evaluating voice quality provided by an embodiment of the present application;

2 is a schematic diagram of a decision tree training classification provided by an embodiment of the present application;

3 is a schematic diagram of another decision tree training provided by an embodiment of the present application;

4 is a flowchart of a method for updating an evaluation model of speech quality provided by an embodiment of the present application;

FIG. 5 is a flowchart of another voice quality evaluation method provided by an embodiment of the present application;

6 is a structural block diagram of an apparatus for evaluating voice quality provided by an embodiment of the present application;

FIG. 7 is a schematic diagram of another apparatus for evaluating voice quality provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

At present, the commonly used voice quality evaluation methods are as follows: extract the parameters of the voice signal or obtain other feature parameters related to voice quality, such as network delay, packet loss, jitter, etc., and then model and analyze the feature parameters to obtain an objective Voice quality evaluation.

Usually, fixed algorithms can be used to model fixed evaluation scenarios, such as the Perceptual evaluation of speech quality (PESQ) algorithm for narrowband speech signals, and the Perceptual Objective Listening for ultra-wideband speech evaluation. QualityAssessment, POLQA) algorithm, etc. The speech quality evaluation models established by the algorithm are all trained linear regression models with specific mapping methods. Finally, the objective speech quality evaluation is mapped with the actual perceived quality of the crowd, and finally the speech quality score is obtained. .

This existing method is suitable for scenarios where the speech environment does not change much. Since the parameters used in training the model are limited, if the speech environment changes greatly, such as on a train, there are many parameters related to speech quality. It may not be limited to the parameters used in the training of the fixed speech quality evaluation model. In this case, the use of this fixed speech quality evaluation model may result in a relatively low accuracy of speech quality evaluation.

In view of this, the embodiment of the present application provides a voice quality evaluation method and device. By continuously acquiring voice signals, the built-in evaluation model is continuously updated based on the input voice signal, and the input voice signal is evaluated, and a voice quality score is output. , thereby improving the accuracy of speech quality evaluation.

It should be understood that words such as "first" and "second" mentioned below are only used for the purpose of distinguishing and describing, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order.

The embodiments of the present application are not constrained by environmental factors, and can be applied to various evaluation environments, including environments with large changes and relatively stable environments.

Secondly, the application scenarios of the embodiments of the present application include but are not limited to traditional second-generation mobile communication technology (2rd-Generation, 2G)/third-generation mobile communication technology (3rd-Generation, 3G) calls, fourth-generation mobile communication technology (the 4th Generation mobile communication technology, 4G) calls, including 2/3/4G mixed scenarios, etc.

FIG. 1 shows a flowchart of a method for evaluating voice quality provided by an embodiment of the present application. Referring to FIG. 1 , the method includes:

S101: Acquire the speech signal to be evaluated, and determine the identification information of the speech signal to be evaluated.

S102: If it is determined that the identification information of the voice signal to be evaluated is different from the identification information of the stored voice signal, the voice signal to be evaluated is regarded as a new voice signal.

It can be understood that the "new voice signal" in the embodiment of the present application means that the voice signal to be evaluated is quite different from the voice signal received before the voice signal to be evaluated is received, and the voice signal to be evaluated can be compared. Mark as new voice signal.

Specifically, the voice quality of the voice signal to be evaluated is different from the voice quality of the stored voice signal, that is, the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are quite different.

Wherein, the voice signal that has been stored in the built-in voice quality evaluation model is the voice signal obtained before the voice signal to be evaluated.

In the embodiment of the present application, the surrounding speech signals are continuously acquired. Therefore, for the speech signal to be evaluated, the speech signal acquired before it can be used as the reference signal of the speech signal to be evaluated.

It should be noted that the voice quality may include, but is not limited to, for example, the voice quality evaluation score and the voice quality level of the voice signal.

S103: When the number of new voice signals is greater than the first preset threshold, update the first voice quality evaluation model to obtain a second voice quality evaluation model.

For the convenience of description, in the embodiments of this application, the "built-in speech quality evaluation model" may be referred to as the "first speech quality evaluation model", and the "updated speech quality evaluation model of the built-in speech quality evaluation model" may be referred to as the "first speech quality evaluation model". Two Speech Quality Evaluation Models".

Specifically, in the embodiment of the present application, when the voice signal to be evaluated is significantly different from the old voice signal, the voice signal to be evaluated can be used as a new sample, and when the number of new samples reaches a preset threshold, for example, it can be the first When there is a preset threshold, the built-in speech quality evaluation model is updated to obtain a second speech quality evaluation model.

It should be noted that in this application, "new voice signal" and "new sample", "old voice signal" and "preserved voice signal" are sometimes used interchangeably, and those skilled in the art should understand that their meanings are consistent .

S104: Use the second speech quality evaluation model to evaluate the speech signal to be evaluated.

Specifically, when the speech signal to be evaluated is a new sample, the updated second speech quality evaluation model can be used to evaluate the speech signal to be evaluated, so as to obtain an accurate speech quality evaluation result.

In the embodiment of the present application, the voice data is continuously obtained through changes in the external voice environment, and the continuously updated data set is used to ensure the accuracy of the voice evaluation model, which can improve the accuracy of the model.

In a possible implementation, the first voice quality evaluation model is updated to obtain a second voice quality evaluation model, which may include:

The characteristic parameters of the new speech signal are acquired, and a decision tree algorithm is used to train the characteristic parameters, and the first speech quality evaluation model is updated to obtain a second speech quality evaluation model.

Specifically, in this embodiment of the present application, a certain number of feature parameters of new voice signals (the number of new voice signals is greater than the first preset threshold) may be extracted, or other feature parameters related to voice quality may be obtained, and then according to the feature parameters A new speech quality evaluation model is obtained by training.

It can be understood that other characteristic parameters related to voice quality include but are not limited to parameters such as network delay, packet loss, and jitter.

The above-mentioned method for obtaining a model by training with characteristic parameters is similar to the existing solution, and will not be repeated here.

In another possible implementation, in this embodiment of the present application, the new voice signal and the old voice signal can be fused, and then the characteristic parameters of the new voice signal and the old voice signal can be extracted, or the characteristic parameters related to the voice quality can be obtained. Other characteristic parameters, and finally a new speech quality evaluation model is obtained by training according to the characteristic parameters.

It should be noted that the old voice signal is the voice signal received before the voice signal to be evaluated.

Specifically, the characteristic parameters in this embodiment of the present application include at least one of the following parameters: signal-to-noise ratio, background noise, noise level, asymmetric interference value of the average speech signal spectrum, high-frequency flatness analysis, spectrum level range, spectrum level Standard Deviation, Relative Noise Floor, Skewness Coefficient of Linear Prediction Coefficient, Cepstral Skewness Coefficient, Voiced Voice, Mean Section of Back Cavity, Vocal Amplitude Variation, Speech Level.

Since there are many speech quality evaluation parameters, some parameters with relatively high weight values are usually selected as characteristic parameters during training.

In this embodiment of the present application, the speech quality evaluation parameters may further include: average speech signal interference value, global background noise, speech interruption time, level dip, silence length, pitch period, mechanization, and the interaction between the back cavity and the middle cavity Relationship, Correlation of Consecutive Frames, Average Power of Consecutive Frames, Energy Sum of Repeated Frames, Number of Repeated Frames, Frames of Unnatural Beeps, Sample Average Energy of Unnatural Beeps, Unnatural Beeps Sample ratio, absolute value of cepstral standard deviation, cepstral kurtosis coefficient, kurtosis coefficient of linear prediction coefficient, absolute value of skewness coefficient of linear prediction coefficient, fixed noise weighting, spectral sharpness, mean energy of samples for background noise level, average energy of samples of background noise, multiplicative noise signal-to-noise ratio, total energy of unnatural silence frames, etc.

Further, after acquiring the speech signal to be evaluated, the method further includes:

The speech signal to be evaluated is subjected to at least one of the following preprocessing: speech data validity detection, speech data normalization processing, and default value difference fitting.

Specifically, there are a large number of incomplete, inconsistent and abnormal data in the original speech signal, which seriously affects the execution efficiency of later modeling, and may even lead to deviations in model results. In addition, the value of the data itself will also affect the results of the model, so data cleaning can be performed on the original speech signal first. It is usually necessary to deal with missing data, anomalies, redundancy, and scaling.

Data processing methods mainly include data validity detection, data normalization, default value interpolation fitting, etc., including but not limited to the above methods.

Further, after acquiring the speech signal to be evaluated, the method further includes:

Evaluate the speech signal to be evaluated according to the first speech quality evaluation model to obtain the speech quality of the speech signal to be evaluated; and classify the speech quality of the speech signal to be evaluated to obtain speech quality of different interval levels.

The voice quality of the different interval levels is used to represent the voice quality of different categories.

There are various options for the classification algorithm in the speech evaluation model, for example, the GBDT (Gradient Boosting Decision Tree, Gradient Boosting Decision Tree) algorithm can be used.

Specifically, in this embodiment of the present application, a decision tree algorithm may be used to classify the quality of the speech signal, as shown in FIG. 2 .

The feature identifiers (1) and (2) in FIG. 2 are used to represent the identification information of the feature parameters of the speech signal. The decision tree algorithm can be regarded as a prediction model, and can also be understood as a classification tree. In this application, a decision tree can be used to map the classification of voice quality.

The classification of speech quality is mapped through a decision tree, and the decision tree can be iterated multiple times to form a progressively improved combination tree to optimize the mapping performance, such as Figure 3. In Figure 3, the learner can predict and score the speech signal. , so as to obtain the predicted speech quality.

The parameters in Figure 3 represent respectively: θ represents the weight, and φ represents the mapping function of different learners.

It should be noted that FIG. 2 and FIG. 3 are only exemplary descriptions, and the specific form and content thereof are not limited to the form and content shown in the figures. For example, the set of scores for speech quality is not limited to classification on a scale of 0-5.

It can be understood that the decision tree can be obtained by methods such as machine learning, which is not limited in this embodiment of the present application.

It can be seen from the advancement algorithm in Figure 3 that the final prediction and scoring result of the speech signal is the combination of the speech quality results of the b learners:

与图中的φ相对应。It is understandable that in the above formula

Corresponds to φ in the figure.

Optimizing the above formula in the function space can be obtained:

where ρ is the learning rate.

According to the above formula, the training value of one speech sample each time can be obtained:

It can be known from the above formula that the voice quality score can correspond to different voice quality score intervals, such as [0, 1], [1, 2], and different voice quality score intervals can correspond to different voice categories.

Preferably, the identification information of the voice signal to be evaluated is different from the identification information of the saved voice signal, and may include:

The speech quality of the speech signal to be evaluated is different from the speech quality of the already stored speech signal, and/or the characteristic parameter of the speech signal to be evaluated is different from the characteristic parameter of the already stored speech signal.

Specifically, the voice quality of the voice signal to be evaluated is different from the voice quality of the voice signal that has been saved, and may include:

The voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are voice signals of the same interval level, and the difference between the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal is less than the set threshold (for example, it may be the second preset threshold), or the voice quality of the voice signal to be evaluated and the voice quality of the voice signal that has been saved are voice signals of different interval levels.

Optionally, in this embodiment of the present application, the built-in speech quality evaluation model (the first speech quality evaluation model) may be used to evaluate the speech signal to be evaluated, so as to determine whether the speech signal to be evaluated is a new speech signal.

Specifically, in the embodiment of the present application, the voice data is obtained from the outside, and the built-in evaluation model is used to classify and score the voice quality of the newly obtained voice data, and then determine whether the newly obtained voice signal belongs to a new sample. If the voice data of different categories are not very different, or the scores of the voice data of the same category differ too much from the scores of the old voice data, this part of the voice data can be used as a new sample.

Specifically, using the characteristic parameter to determine that the identification information of the speech signal to be evaluated is different from the identification information of the stored speech signal may include but not be limited to the following methods:

(1), based on the univariate normal distribution detection:

The original data set is _xi,1 , _xi,2 , _xi,3 ,..., _xi,n ,i∈(1,...,m), including m samples, n-dimensional features, each feature can be calculated The mean and method of dimensions:

For new data, the probability can be calculated as:

The difference in feature distribution between the new data and the old data can be judged according to the probability.

(2), based on multivariate Gaussian distribution detection:

共n维特征向量，可以计算n维特征均值向量和n*n的协方差矩阵：The original data set is

For a total of n-dimensional eigenvectors, the n-dimensional eigenmean vector and the n*n covariance matrix can be calculated:

Σ=[Cov(x _i , x _j )], i, j∈(1,...,n)

For new data, the probability can be calculated as:

The difference between the feature distribution of the new data and the old data can be judged according to the probability, where T in the formula represents the transposition of the matrix.

(3), based on Mahalanobis distance detection:

For multidimensional data sets, a is the mean vector, and the Mahalanobis distance from the new data a to a is:

Among them, T represents the transpose of the matrix, and S is the covariance matrix. If the value of S is too large, the feature distribution is considered to be different.

(4), based on feature importance detection:

The importance ranking of features can be derived using tree-based models such as GBDT.

The global importance of feature j is measured by the average of the importance of feature j in a single tree:

where M is the number of trees.

The importance of feature j in a single tree is as follows:

是节点t分裂后平方损失减少值，J表示特征集合，T表示树的集合。Among them, L is the number of leaf nodes of the tree, and L-1 is the number of non-leaf nodes of the tree. ν _t is the feature associated with node t,

is the squared loss reduction value after node t is split, J represents the feature set, and T represents the set of trees.

For the first k features trained by the new sample, if they are quite different from the features of the original dataset, it is considered that the distribution is different from the original dataset.

In the embodiment of the present application, in a possible implementation manner, the speech data may be incrementally learned through the method flow chart shown in FIG. 4 , thereby updating the built-in speech quality evaluation model, as shown in FIG. 4 .

It can be understood that the normal scoring in FIG. 4 is the scoring performed by the built-in speech quality evaluation model.

For the entire method flow in the embodiment of the present application, the method flow chart shown in FIG. 5 can be involved. In this method, by acquiring an external voice signal, preprocessing the voice signal, and then using a decision tree algorithm to perform voice Classify the signal quality, get the quality score of the voice signal, and then judge whether the voice sample data conforms to the new sample characteristics. When the voice signal is a new sample, after collecting a certain number of new samples, the built-in voice quality evaluation model is updated. Standard scoring is performed using the updated speech quality evaluation model.

Based on the same concept as the above method embodiment, an embodiment of the present invention also provides a structural block diagram of an apparatus for evaluating voice quality. Referring to FIG. 6 , the apparatus includes: an acquisition unit 101 , a determination unit 102 , and an update unit 103. Evaluation unit 104.

Wherein, the obtaining unit 101 is configured to obtain the speech signal to be evaluated.

The determining unit 102 is configured to determine the identification information of the voice signal to be evaluated obtained by the acquiring unit 101, and when determining that the voice quality of the voice signal to be evaluated is different from the voice quality of the stored voice signal, the voice signal as the new voice signal.

The updating unit 103 updates the first speech quality evaluation model to obtain a second speech quality evaluation model when the number of new speech signals determined by the determining unit 102 is greater than the first preset threshold.

The stored voice signal is the voice signal obtained before the voice signal to be evaluated.

The evaluating unit 104 is configured to use the second speech quality evaluation model obtained by the updating unit 103 to evaluate the speech signal to be evaluated.

Specifically, the updating unit 103 is specifically configured to update the first voice quality evaluation model in the following manner to obtain a second voice quality evaluation model:

Obtain the characteristic parameters of the new speech signal; use the decision tree algorithm to train the characteristic parameters, update the first speech quality evaluation model, and obtain the second speech quality evaluation model.

The characteristic parameters include at least one of the following: signal-to-noise ratio, background noise, noise level, asymmetric interference value of the average speech signal spectrum, high-frequency flatness analysis, spectrum level range, spectrum level standard deviation, relative noise floor, linearity Skewness coefficient of prediction coefficients, cepstral skewness coefficient, voiced sound, mean cross section of back cavity, vocal tract amplitude variation, speech level.

Correspondingly, the apparatus further includes: the processing unit 105 is configured to:

At least one of the following preprocessing is performed on the speech signal to be evaluated: speech data validity detection, speech data normalization processing, and default value difference fitting.

Further, the evaluation unit 104 is also used for:

The speech signal to be evaluated is evaluated according to the first speech quality evaluation model to obtain the speech quality of the speech signal to be evaluated.

The processing unit 105 is also used for:

The speech quality of the speech signal to be evaluated is classified to obtain speech quality of different interval levels; the speech quality of different interval levels is used to characterize the speech quality of different categories.

Optionally, the identification information of the voice signal to be evaluated is different from the identification information of the saved voice signal, including:

The speech quality of the speech signal to be evaluated is different from the speech quality of the already stored speech signal, and/or the characteristic parameter of the speech signal to be evaluated is different from the characteristic parameter of the already stored speech signal.

Further, the voice quality of the voice signal to be evaluated is different from the voice quality of the voice signal that has been saved, including:

The voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are voice signals of the same interval level, and the difference between the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal is less than the first Two preset thresholds; or the voice quality of the voice signal to be evaluated and the voice quality of the saved voice signal are voice signals of different interval levels.

It should be noted that, for the function implementation of each unit in the above-mentioned apparatus for evaluating voice quality in the embodiments of the present invention, further reference may be made to the description of the related method embodiments, and details are not repeated here.

The embodiment of the present application also provides another voice quality evaluation device, as shown in FIG. 7 , the device includes:

Memory 202 for storing program instructions.

The transceiver 201 is used for receiving and sending voice quality evaluation instructions.

The processor 200 is configured to call the program instructions stored in the memory, and execute the processing unit (102), the determining unit (103), the determining unit (103), Methods performed by the update unit (104) and the evaluation unit (105).

7, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 200 and various circuits of memory represented by memory 202 are linked together. The bus architecture may also link together various other circuits, such as peripherals, voltage regulators, and power management circuits, which are well known in the art and, therefore, will not be described further herein. The bus interface provides the interface.

Transceiver 201 may be a number of elements, including a transmitter and a transceiver, providing a means for communicating with various other devices over a transmission medium.

The processor 200 is responsible for managing the bus architecture and general processing, and the memory 202 may store data used by the processor 200 in performing operations.

The processor 200 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or a complex programmable logic device (CPLD) ).

The embodiments of the present application further provide a computer storage medium for storing computer program instructions used by any of the apparatuses described in the above embodiments of the present application, including instructions for executing any of the methods provided by the above embodiments of the present application program of.

The computer storage medium can be any available medium or data storage device that can be accessed by a computer, including but not limited to magnetic storage (eg, floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical storage (eg CD, DVD, BD, HVD, etc.), and semiconductor memory (eg, ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state disk (SSD)), and the like.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be construed to include the preferred embodiment and all changes and modifications that fall within the scope of the present invention.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention. Thus, provided that these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims (14)

1. A method for evaluating voice quality, comprising:

acquiring a voice signal to be evaluated, and determining identification information of the voice signal to be evaluated;

if the identification information of the voice signal to be evaluated is different from the identification information of the stored voice signal, taking the voice signal to be evaluated as a new voice signal, and updating the first voice quality evaluation model when the number of the new voice signals is larger than a first preset threshold value to obtain a second voice quality evaluation model;

wherein the stored voice signal is a voice signal acquired before the voice signal to be evaluated;

and evaluating the voice signal to be evaluated by utilizing the second voice quality evaluation model.

2. The method of claim 1, wherein updating the first speech quality assessment model to obtain a second speech quality assessment model comprises:

acquiring the characteristic parameters of the new voice signal;

training the characteristic parameters by using a decision tree algorithm, and updating the first voice quality evaluation model to obtain a second voice quality evaluation model;

the characteristic parameters include at least one of: signal-to-noise ratio, background noise, noise level, asymmetric interference value of average speech signal spectrum, high-frequency flatness analysis, spectrum level range, spectrum level standard deviation, relative noise floor, skewness coefficient of linear prediction coefficient, cepstrum skewness coefficient, voiced sound, average cross section of back cavity, vocal tract amplitude variation and speech level.

3. The method of claim 1, wherein after acquiring the speech signal to be evaluated, the method further comprises:

performing at least one of the following preprocessing on the voice signal to be evaluated: voice data validity detection, voice data normalization processing and default value difference fitting.

4. The method of claim 1, wherein after acquiring the speech signal to be evaluated, the method further comprises:

evaluating the voice signal to be evaluated according to the first voice quality evaluation model to obtain the voice quality of the voice signal to be evaluated;

classifying the voice quality of the voice signal to be evaluated to obtain the voice quality of different interval grades; the speech quality of the different interval levels is used for characterizing the speech quality of different classes.

5. The method of claim 1, wherein the identification information of the speech signal to be evaluated is different from the identification information of the speech signal already stored, comprising:

the speech quality of the speech signal to be evaluated differs from the speech quality of the already stored speech signal and/or the characteristic parameters of the speech signal to be evaluated differ from the characteristic parameters of the already stored speech signal.

6. The method of claim 5, wherein the speech quality of the speech signal to be evaluated is different from the speech quality of the speech signal already stored, comprising:

the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are voice signals of the same interval grade, and the difference value between the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal is smaller than a second preset threshold value; or

And the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are voice signals with different interval grades.

7. An apparatus for evaluating speech quality, comprising:

the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a voice signal to be evaluated;

the determining unit is used for determining the identification information of the voice signal to be evaluated, and taking the voice signal to be evaluated as a new voice signal when the voice quality of the voice signal to be evaluated is determined to be different from the voice quality of the stored voice signal;

the updating unit is used for updating the first voice quality evaluation model to obtain a second voice quality evaluation model when the number of the new voice signals is larger than a first preset threshold value;

the stored voice signal is a voice signal which is evaluated by the first voice quality evaluation model before the voice signal to be evaluated;

and the evaluation unit is used for evaluating the voice signal to be evaluated by utilizing the second voice quality evaluation model.

8. The apparatus according to claim 7, wherein the updating unit is specifically configured to update the first speech quality assessment model to obtain a second speech quality assessment model as follows:

acquiring the characteristic parameters of the new voice signal;

training the characteristic parameters by using a decision tree algorithm, and updating the first voice quality evaluation model to obtain a second voice quality evaluation model;

the characteristic parameters include at least one of: signal-to-noise ratio, background noise, noise level, asymmetric interference value of average speech signal spectrum, high-frequency flatness analysis, spectrum level range, spectrum level standard deviation, relative noise floor, skewness coefficient of linear prediction coefficient, cepstrum skewness coefficient, voiced sound, average cross section of back cavity, vocal tract amplitude variation and speech level.

9. The apparatus of claim 7, further comprising a processing unit to:

performing at least one of the following preprocessing on the voice signal to be evaluated: voice data validity detection, voice data normalization processing and default value difference fitting.

10. The apparatus of claim 7, wherein the evaluation unit is further to:

evaluating the voice signal to be evaluated according to the first voice quality evaluation model to obtain the voice quality of the voice signal to be evaluated;

the processing unit is further to:

classifying the voice quality of the voice signal to be evaluated to obtain the voice quality of different interval grades; the speech quality of the different interval levels is used for characterizing the speech quality of different classes.

11. The apparatus of claim 7, wherein the identification information of the speech signal to be evaluated is different from the identification information of the speech signal already stored, comprising:

the speech quality of the speech signal to be evaluated differs from the speech quality of the already stored speech signal and/or the characteristic parameters of the speech signal to be evaluated differ from the characteristic parameters of the already stored speech signal.

12. The apparatus of claim 11, wherein the speech quality of the speech signal to be evaluated is different from the speech quality of the speech signal already stored, comprising:

the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are voice signals of the same interval grade, and the difference value between the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal is smaller than a second preset threshold value; or

And the voice quality of the voice signal to be evaluated and the voice quality of the stored voice signal are voice signals with different interval grades.

13. An apparatus for evaluating speech quality, comprising:

a memory for storing program instructions;

a processor for calling the program instructions stored in the memory and executing the method of any one of claims 1 to 6 according to the obtained program.

14. A computer readable storage medium having stored thereon computer instructions which, when run on a computer, cause the computer to perform the method of any of claims 1-6.

Images