A piano data evaluation system and method based on machine learning

Abstract

The present invention relates to the technical field of audio data processing, and in particular to a piano data evaluation system and method based on machine learning. The method comprises: obtaining piano audio data, and obtaining actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data; calculating the audio spectrum distortion based on the actual load amplitude response and the audio receiving frequency, calculating the audio quality index based on the audio evaluation quantity, and calculating the loading coefficient JZ based on the spectrum distortion and the audio quality index, and judging whether the piano audio data meets the evaluation requirements by the loading coefficient JZ, and loading the piano audio data that meets the evaluation into the evaluation module; receiving the piano audio data and evaluating it, obtaining the evaluation score and the historical evaluation score, and generating an evaluation report. The present invention can reduce the subjectivity of the evaluation, improve the evaluation efficiency, and reduce the evaluation cost when evaluating the piano audio.

Description

Technical Field

[0001] The present invention relates to the technical field of audio data processing, and in particular to a piano data evaluation system and method based on machine learning. Background Art

[0002] With the rapid development of computer technology and artificial intelligence, machine learning technology has been widely used in many fields such as music creation, performance analysis, and music education. In terms of piano audio evaluation, the traditional manual evaluation method has problems such as strong subjectivity, low efficiency, and high cost, which cannot meet the needs of modern music education for objective and efficient evaluation. Therefore, developing a piano data evaluation system based on machine learning has become an important way to solve the above problems.

[0003] At present, there are some piano audio evaluation systems based on computer technology on the market, but most of them have the following shortcomings: Single evaluation dimension: only focusing on a single dimension such as intonation or rhythm, it cannot fully reflect the overall level of the performer. Strong subjectivity: some systems still rely on manually set rules and thresholds, and the evaluation results are easily affected by subjective factors.

[0004] Therefore, there is an urgent need for a piano data evaluation system and method based on machine learning to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a piano data evaluation system and method based on machine learning: to solve the technical problems that the evaluation method of piano audio in the existing scheme is highly subjective, inefficient and costly.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] On the one hand, a piano data evaluation system based on machine learning includes a piano audio data acquisition module, a piano audio data loading module, an evaluation module and a storage module;

[0008] The piano audio data acquisition module is used to acquire piano audio data, and obtain the actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data;

[0009] The piano audio data loading module is used to calculate the audio spectrum distortion based on the actual load amplitude response and the audio receiving frequency, calculate the audio quality index based on the audio evaluation quantity, and calculate the loading coefficient JZ based on the spectrum distortion and the audio quality index, and judge whether the piano audio data meets the evaluation requirements through the loading coefficient JZ, and load the piano audio data that meets the evaluation into the evaluation module;

[0010] The evaluation module is used to receive piano audio data and perform evaluation: preprocess the piano audio data, and perform noise estimation on the preprocessed piano audio data to obtain a noise spectrum, obtain a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculate a gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio, and enhance the piano audio data based on the gain factor, and substitute the enhanced piano audio data into the evaluation model based on the neural network to obtain an evaluation score;

[0011] The storage module is used to obtain evaluation scores and historical evaluation scores and generate evaluation reports.

[0012] Furthermore, obtaining the actual load amplitude response based on the piano audio data specifically includes the following process:

[0013] A detection collection point is set, and at a preset audio receiving frequency, the amplitude response of the piano audio data at the detection collection point is detected in real time, a rectangular coordinate system is established with the detection execution time of the detection collection point as the X-axis and the amplitude response as the Y-axis, and an actual load amplitude response curve is drawn in the rectangular coordinate system through the amplitude response value, and two truncated line segments are obtained by drawing perpendicular lines from the two end points of the actual load amplitude response to the X-axis. A figure is formed by the actual load amplitude response curve, the two truncated line segments and the X-axis, and the area value of the figure is marked as the actual load amplitude response.

[0014] Furthermore, the audio spectrum distortion is calculated based on the actual load amplitude response and the audio receiving frequency, which specifically includes the following process:

[0015] Loading table load amplitude response L(f g ), where the load amplitude response L(f g ) is stored in a storage module;

[0016] The actual load amplitude response H(f g )、Table load amplitude response L(f g ) is substituted into the audio spectrum distortion calculation formula to obtain the audio spectrum distortion SD, and the audio spectrum distortion calculation formula is as follows:

[0017]

[0018] in, Indicates taking The absolute value of , G represents the number of audio evaluations.

[0019] Furthermore, the audio quality index is calculated based on the number of audio evaluations and specifically includes the following process:

[0020] Get the quality score weights σ corresponding to pitch, rhythm and timbre i , where 0<i<4, i is an integer;

[0021] The quality score weight σi Substitute the audio evaluation quantity G into the audio quality index calculation formula to obtain the audio quality index LMO, which is calculated as follows:

[0022] LMO=(∑ 0＜g＜G score g ) / G;

[0023] score g =(∑ 0＜i＜4 σ i *(Z)) / G;

[0024] Among them, Z is the comprehensive value of pitch, rhythm and timbre corresponding to the g-th audio, and the comprehensive value is obtained by adding the pitch value, rhythm energy value and the number of timbre harmonic components. Among them, the sound frequency analyzer measures the pitch value; the Audacity editing software measures the energy value; and the spectrum analyzer detects the number of harmonic components of the timbre in the piano audio.

[0025] Furthermore, the loading coefficient JZ is calculated based on the spectrum distortion and the audio quality index, and the loading coefficient JZ is used to determine whether the piano audio data meets the evaluation requirements. Specifically, the process includes the following steps:

[0026] Subtract the spectral distortion from the audio quality index to obtain the loading coefficient JZ;

[0027] Obtaining a loading coefficient threshold, wherein the loading coefficient threshold is stored in a storage module;

[0028] It is determined whether the loading coefficient JZ exceeds the loading coefficient threshold. If so, the piano audio data meets the evaluation requirements. If not, the piano audio data does not meet the evaluation requirements.

[0029] Further, the piano audio data is preprocessed, and the noise of the preprocessed piano audio data is estimated to obtain the noise spectrum, which specifically includes the following process:

[0030] Preprocessing the piano audio data: dividing the piano audio data into frames and adding windows;

[0031] Perform Fourier transform on the framed and windowed piano audio data:

[0032]

[0033] Where s represents the noisy speech stream, n represents the nth sample point of the mth frame of the piano audio data, K represents the number of samples in a frame, c(k) represents the frame windowing process, L represents the displacement of each frame, and e -2p(π / K)kn represents the Fourier transform coefficient; D(n, m) is the short-time spectrum of the original speech;

[0034] Perform compensation processing on D(n, m) to obtain the compensated original speech spectrum T(n, m):

[0035] T(n,m)=T(n,m-1)*α+(1-α)*D(n,m) 2 ;

[0036] Among them, α is a fixed compensation factor;

[0037] Scan T(n, m), search the current frame and the frames before and after the current frame, obtain the smallest spectral component of the previous frames and the smallest spectral component of the next frames, take the maximum value of these two values ​​as the current reference value, and calculate the speech existence probability P(n, m) based on the current reference value:

[0038]

[0039] Among them, T s (n, m) is the current reference value, β is the fixed coefficient, and μ is the threshold coefficient;

[0040] The noise spectrum d(n, m) is calculated based on P(n, m):

[0041]

[0042]

[0043] Among them, ω is the compensation factor and γ is the constant coefficient.

[0044] Further, obtaining a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculating the gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio specifically include the following processes:

[0045] Obtain a noise power spectrum based on the noise spectrum, and obtain a piano audio power spectrum, and record the ratio of the piano audio power spectrum to the noise power spectrum as a priori signal-to-noise ratio;

[0046] The ratio of the power spectrum of the piano audio power spectrum after noise reduction to the noise power spectrum is recorded as the posterior signal-to-noise ratio:

[0047] The gain factor G(n, m) is calculated based on the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio:

[0048] G(n,m)=X(n,m) / (1-H(n,m))*θ;

[0049] Among them, X(n, m) is the prior signal-to-noise ratio, H(n, m) is the posterior signal-to-noise ratio, and θ is the gain coefficient.

[0050] Furthermore, performing piano audio data enhancement according to the gain factor specifically includes the following process: multiplying the gain factor by the piano audio data to obtain enhanced audio data, and then performing inverse Fourier transformation on the enhanced audio data and then superimposing them.

[0051] Furthermore, the enhanced piano audio data is substituted into the evaluation model based on the neural network to obtain the evaluation score, which specifically includes the following process:

[0052] Feature extraction: converting piano audio data into feature representation suitable for neural network processing, the feature representation is a spectrogram;

[0053] Data preparation: Divide the preprocessed piano audio data into training set, validation set and test set;

[0054] Model design: Design a suitable network structure according to task requirements, including the number of layers, number of neurons, and activation function;

[0055] Training and optimization: Use training data to train the model and adjust network parameters through the back-propagation algorithm to minimize the loss function;

[0056] Evaluation and adjustment: Use the validation set to evaluate model performance and adjust the model structure and parameters based on the evaluation results;

[0057] Get the evaluation score of the piano audio data based on the preset labels.

[0058] On the other hand, a piano data evaluation method based on machine learning is applicable to the piano data evaluation system based on machine learning described above, and the method comprises:

[0059] Acquire piano audio data, and obtain actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data;

[0060] The audio spectrum distortion is calculated based on the actual load amplitude response and the audio receiving frequency, the audio quality index is calculated based on the audio evaluation quantity, and the loading coefficient JZ is calculated based on the spectrum distortion and the audio quality index. Whether the piano audio data meets the evaluation requirements is determined by the loading coefficient JZ, and the piano audio data that meets the evaluation requirements is loaded into the evaluation module;

[0061] Receiving piano audio data and conducting evaluation: preprocessing the piano audio data, and performing noise estimation on the preprocessed piano audio data to obtain a noise spectrum, obtaining a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculating a gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio, and enhancing the piano audio data based on the gain factor, substituting the enhanced piano audio data into an evaluation model based on a neural network to obtain an evaluation score;

[0062] Get evaluation scores and historical evaluation scores, and generate evaluation reports.

[0063] Compared with the existing solutions, the present invention achieves the following beneficial effects:

[0064] The present invention obtains piano audio data, and obtains actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data; calculates the audio spectrum distortion based on the actual load amplitude response and the audio receiving frequency, calculates the audio quality index based on the audio evaluation quantity, and calculates the loading coefficient JZ based on the spectrum distortion and the audio quality index, and determines whether the piano audio data meets the evaluation requirements through the loading coefficient JZ, and loads the piano audio data that meets the evaluation into the evaluation module; receives the piano audio data and evaluates it, obtains the evaluation score and the historical evaluation score, and generates an evaluation report. The present invention can reduce the subjectivity of evaluation, improve the evaluation efficiency, and reduce the evaluation cost when evaluating piano audio. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0066] Figure 1 is a system block diagram of a piano data evaluation system based on machine learning according to an embodiment of the present invention;

[0067] Figure 2 is a work flow chart of a piano data evaluation system based on machine learning according to an embodiment of the present invention;

[0068] Figure 3 is another work flow chart of a piano data evaluation system based on machine learning according to an embodiment of the present invention;

[0069] Figure 4 This is a workflow diagram of a piano data evaluation method based on machine learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0072] This embodiment provides a piano data evaluation system based on machine learning. Figure 1 is a system block diagram of a piano data evaluation system based on machine learning according to an embodiment of the present invention, such as Figure 1 As shown, the system includes a piano audio data acquisition module, a piano audio data loading module, an evaluation module and a storage module;

[0073] The piano audio data acquisition module is used to acquire piano audio data, and obtain the actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data;

[0074] The piano audio data loading module is used to calculate the audio spectrum distortion based on the actual load amplitude response and the audio receiving frequency, calculate the audio quality index based on the audio evaluation quantity, and calculate the loading coefficient JZ based on the spectrum distortion and the audio quality index, and judge whether the piano audio data meets the evaluation requirements through the loading coefficient JZ, and load the piano audio data that meets the evaluation into the evaluation module;

[0075] The evaluation module is used to receive piano audio data and perform evaluation: preprocess the piano audio data, and perform noise estimation on the preprocessed piano audio data to obtain a noise spectrum, obtain a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculate a gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio, and enhance the piano audio data based on the gain factor, and substitute the enhanced piano audio data into the evaluation model based on the neural network to obtain an evaluation score;

[0076] The storage module is used to obtain evaluation scores and historical evaluation scores and generate evaluation reports.

[0077] In summary, the present invention obtains piano audio data, and obtains the actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data; the audio spectrum distortion is calculated based on the actual load amplitude response and the audio receiving frequency, the audio quality index is calculated based on the audio evaluation quantity, and the loading coefficient JZ is calculated based on the spectrum distortion and the audio quality index, and the piano audio data is judged by the loading coefficient JZ whether the evaluation requirements are met, and the piano audio data that meets the evaluation is loaded into the evaluation module; the piano audio data is received and evaluated, the evaluation score and the historical evaluation score are obtained, and an evaluation report is generated. The present invention can reduce the subjectivity of evaluation, improve the evaluation efficiency, and reduce the evaluation cost when evaluating piano audio.

[0078] In some embodiments, obtaining the actual load amplitude response based on the piano audio data specifically includes the following process:

[0079] A detection collection point is set, and at a preset audio receiving frequency, the amplitude response of the piano audio data at the detection collection point is detected in real time, a rectangular coordinate system is established with the detection execution time of the detection collection point as the X-axis and the amplitude response as the Y-axis, and an actual load amplitude response curve is drawn in the rectangular coordinate system through the amplitude response value, and two truncated line segments are obtained by drawing perpendicular lines from the two end points of the actual load amplitude response to the X-axis. A figure is formed by the actual load amplitude response curve, the two truncated line segments and the X-axis, and the area value of the figure is marked as the actual load amplitude response.

[0080] In some embodiments, calculating the audio spectrum distortion based on the actual load amplitude response and the audio reception frequency specifically includes the following process:

[0081] Loading table load amplitude response L(f g ), where the load amplitude response L(f g ) is stored in a storage module;

[0082] The actual load amplitude response H(f g )、Table load amplitude response L(f g ) is substituted into the audio spectrum distortion calculation formula to obtain the audio spectrum distortion SD, and the audio spectrum distortion calculation formula is as follows:

[0083]

[0084] in, Indicates taking The absolute value of , G represents the number of audio evaluations.

[0085] In some embodiments, calculating the audio quality index based on the number of audio evaluations specifically includes the following process:

[0086] Get the quality score weights σ corresponding to pitch, rhythm and timbre i, where 0<i<4, i is an integer;

[0087] The quality score weight σ i Substitute the audio evaluation quantity G into the audio quality index calculation formula to obtain the audio quality index LMO, which is calculated as follows:

[0088] LMO=(∑ 0＜g＜G score g ) / G;

[0089] score g =(∑ 0＜i＜4 σ i *(Z)) / G;

[0090] Among them, Z is the comprehensive value of pitch, rhythm and timbre corresponding to the g-th audio, and the comprehensive value is obtained by adding the pitch value, rhythm energy value and the number of timbre harmonic components. Among them, the sound frequency analyzer measures the pitch value; the Audacity editing software measures the energy value; and the spectrum analyzer detects the number of harmonic components of the timbre in the piano audio.

[0091] In some embodiments, Figure 2 is a work flow chart of a piano data evaluation system based on machine learning according to an embodiment of the present invention, such as Figure 2 As shown, the loading coefficient JZ is calculated based on the spectrum distortion and the audio quality index, and the loading coefficient JZ is used to determine whether the piano audio data meets the evaluation requirements, which specifically includes the following steps:

[0092] Step S201: Subtract the spectral distortion from the audio quality index to obtain a loading coefficient JZ;

[0093] Step S202: obtaining a loading coefficient threshold, wherein the loading coefficient threshold is stored in a storage module;

[0094] Step S203: determine whether the loading coefficient JZ exceeds the loading coefficient threshold, if so, proceed to step S204, if not, proceed to step S205;

[0095] Step S204: The piano audio data meets the evaluation requirements;

[0096] Step S205: The piano audio data does not meet the evaluation requirements.

[0097] In some embodiments, preprocessing the piano audio data and performing noise estimation on the preprocessed piano audio data to obtain the noise spectrum specifically includes the following process:

[0098] Preprocessing the piano audio data: dividing the piano audio data into frames and adding windows;

[0099] Perform Fourier transform on the framed and windowed piano audio data:

[0100]

[0101] Where s represents the noisy speech stream, n represents the nth sample point of the mth frame of the piano audio data, K represents the number of samples in a frame, c(k) represents the frame windowing process, L represents the displacement of each frame, and e -2p(π / K)kn represents the Fourier transform coefficient; D(n, m) is the short-time spectrum of the original speech;

[0102] Perform compensation processing on D(n, m) to obtain the compensated original speech spectrum T(n, m):

[0103] T(n,m)=T(n,m-1)*α+(1-α)*D(n,m)2;

[0104] Among them, α is a fixed compensation factor;

[0105] Scan T(n, m), search the current frame and the frames before and after the current frame, obtain the smallest spectral component of the previous frames and the smallest spectral component of the next frames, take the maximum value of these two values ​​as the current reference value, and calculate the speech existence probability P(n, m) based on the current reference value:

[0106]

[0107] Among them, T s (n, m) is the current reference value, β is the fixed coefficient, and μ is the threshold coefficient;

[0108] The noise spectrum d(n, m) is calculated based on P(n, m):

[0109]

[0110]

[0111] Among them, ω is the compensation factor and γ is the constant coefficient.

[0112] In some embodiments, Figure 3 FIG. 4 is another work flow chart of a piano data evaluation system based on machine learning according to an embodiment of the present invention. Figure 3 As shown, obtaining a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculating the gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio specifically include the following steps:

[0113] Step S301: obtaining a noise power spectrum based on the noise spectrum, and obtaining a piano audio power spectrum, and recording the ratio of the piano audio power spectrum to the noise power spectrum as a priori signal-to-noise ratio;

[0114] Step S302: Record the ratio of the power spectrum of the piano audio after the noise reduction process to the noise power spectrum as the posterior signal-to-noise ratio;

[0115] Step S303: Calculate a gain factor based on the a priori signal-to-noise ratio and the a posteriori signal-to-noise ratio.

[0116] Specifically, the gain factor G(n, m) is calculated based on the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio:

[0117] G(n,m)=x(n,m) / (1-H(n,m))*θ;

[0118] Among them, X(n, m) is the prior signal-to-noise ratio, H(n, m) is the posterior signal-to-noise ratio, and θ is the gain coefficient.

[0119] In some embodiments, performing piano audio data enhancement according to a gain factor specifically includes the following process: multiplying the gain factor by the piano audio data to obtain enhanced audio data, and then performing an inverse Fourier transform on the enhanced audio data and then superimposing the data.

[0120] Furthermore, the enhanced piano audio data is substituted into the evaluation model based on the neural network to obtain the evaluation score, which specifically includes the following process:

[0121] Feature extraction: converting piano audio data into feature representation suitable for neural network processing, the feature representation is a spectrogram;

[0122] Data preparation: Divide the preprocessed piano audio data into training set, validation set and test set;

[0123] Model design: Design a suitable network structure according to task requirements, including the number of layers, number of neurons, and activation function;

[0124] Training and optimization: Use training data to train the model and adjust network parameters through the back-propagation algorithm to minimize the loss function;

[0125] Evaluation and adjustment: Use the validation set to evaluate model performance and adjust the model structure and parameters based on the evaluation results;

[0126] Get the evaluation score of the piano audio data based on the preset labels.

[0127] On the other hand, the present invention also provides a piano data evaluation method based on machine learning, which is applicable to any of the above-mentioned commodity logistics sorting management systems based on artificial intelligence. Figure 4 is a workflow diagram of a piano data evaluation method based on machine learning according to an embodiment of the present invention, such as Figure 4 As shown, the method comprises the following steps:

[0128] Step S401: Acquire piano audio data, and obtain actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data;

[0129] Step S402: Calculate the audio spectrum distortion based on the actual load amplitude response and the audio receiving frequency, calculate the audio quality index based on the audio evaluation quantity, and calculate the loading coefficient JZ based on the spectrum distortion and the audio quality index, and determine whether the piano audio data meets the evaluation requirements through the loading coefficient JZ, and load the piano audio data that meets the evaluation requirements into the evaluation module;

[0130] Step S403: receiving piano audio data and performing evaluation: preprocessing the piano audio data, and performing noise estimation on the preprocessed piano audio data to obtain a noise spectrum, obtaining a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculating a gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio, and enhancing the piano audio data based on the gain factor, substituting the enhanced piano audio data into an evaluation model based on a neural network to obtain an evaluation score;

[0131] Step S404: Obtain evaluation scores and historical evaluation scores, and generate an evaluation report.

[0132] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0133] 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, or a combination of computer software and electronic hardware. 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 application.

[0134] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0135] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. 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. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0136] 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 solution of this embodiment.

[0137] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A piano data evaluation system based on machine learning, characterized in that: The system includes a piano audio data acquisition module, a piano audio data loading module, an evaluation module and a storage module; The piano audio data acquisition module is used to acquire piano audio data, and obtain the actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data; The piano audio data loading module is used to calculate the audio spectrum distortion based on the actual load amplitude response and the audio receiving frequency, calculate the audio quality index based on the audio evaluation quantity, and calculate the loading coefficient JZ based on the spectrum distortion and the audio quality index, and judge whether the piano audio data meets the evaluation requirements through the loading coefficient JZ, and load the piano audio data that meets the evaluation into the evaluation module; The evaluation module is used to receive piano audio data and perform evaluation: preprocess the piano audio data, and perform noise estimation on the preprocessed piano audio data to obtain a noise spectrum, obtain a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculate a gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio, and enhance the piano audio data based on the gain factor, and substitute the enhanced piano audio data into the evaluation model based on the neural network to obtain an evaluation score; The storage module is used to obtain evaluation scores and historical evaluation scores and generate evaluation reports.

2. A piano data evaluation system based on machine learning according to claim 1, characterized in that: The specific process of obtaining the actual load amplitude response based on piano audio data includes the following steps: A detection collection point is set, and at a preset audio receiving frequency, the amplitude response of the piano audio data at the detection collection point is detected in real time, a rectangular coordinate system is established with the detection execution time of the detection collection point as the X-axis and the amplitude response as the Y-axis, and an actual load amplitude response curve is drawn in the rectangular coordinate system through the amplitude response value, and two truncated line segments are obtained by drawing perpendicular lines from the two end points of the actual load amplitude response to the X-axis. A figure is formed by the actual load amplitude response curve, the two truncated line segments and the X-axis, and the area value of the figure is marked as the actual load amplitude response.

3. A piano data evaluation system based on machine learning according to claim 1, characterized in that: The audio spectrum distortion is calculated based on the actual load amplitude response and the audio receiving frequency, which specifically includes the following process: Loading table load amplitude response L(f g ), where the load amplitude response L(f g ) is stored in a storage module; The actual load amplitude response H(f g )、Table load amplitude response L(f g ) is substituted into the audio spectrum distortion calculation formula to obtain the audio spectrum distortion SD, and the audio spectrum distortion calculation formula is as follows: in, Indicates taking The absolute value of , G represents the number of audio evaluations.

4. A piano data evaluation system based on machine learning according to claim 1, characterized in that: The specific process of calculating the audio quality index based on the number of audio evaluations is as follows: Get the quality score weights σ corresponding to pitch, rhythm and timbre i , where 0<i<4, i is an integer; The quality score weight σ i Substitute the audio evaluation quantity G into the audio quality index calculation formula to obtain the audio quality index LMO, which is calculated as follows: LMO=(∑ 0＜g <Gscoreg) / G; score g =(∑ 0<i＜4 σ i *(Z)) / G; Among them, Z is the comprehensive value of pitch, rhythm and timbre corresponding to the g-th audio, and the comprehensive value is obtained by adding the pitch value, rhythm energy value and the number of timbre harmonic components. Among them, the sound frequency analyzer measures the pitch value; the Audacity editing software measures the energy value; and the spectrum analyzer detects the number of harmonic components of the timbre in the piano audio.

5. A piano data evaluation system based on machine learning according to claim 1, characterized in that: The loading coefficient JZ is calculated based on the spectrum distortion and the audio quality index, and the loading coefficient JZ is used to determine whether the piano audio data meets the evaluation requirements. The specific process includes the following: Subtract the spectral distortion from the audio quality index to obtain the loading coefficient JZ; Obtaining a loading coefficient threshold, wherein the loading coefficient threshold is stored in a storage module; It is determined whether the loading coefficient JZ exceeds the loading coefficient threshold. If so, the piano audio data meets the evaluation requirements. If not, the piano audio data does not meet the evaluation requirements.

6. A piano data evaluation system based on machine learning according to claim 1, characterized in that: The piano audio data is preprocessed, and the noise of the preprocessed piano audio data is estimated to obtain the specific noise spectrum. The process includes: Preprocessing the piano audio data: dividing the piano audio data into frames and adding windows; Perform Fourier transform on the framed and windowed piano audio data: Where s represents the noisy speech stream, n represents the nth sample point of the mth frame of the piano audio data, K represents the number of samples in a frame, c(k) represents the frame windowing process, L represents the displacement of each frame, and e -2p(π / K)kn represents the Fourier transform coefficient; D(n, m) is the short-time spectrum of the original speech; Perform compensation processing on D(n, m) to obtain the compensated original speech spectrum T(n, m): T(n,m)=T(n,m-1)*α+(1-α)*D(n,m) 2 ; Among them, α is a fixed compensation factor; Scan T(n, m), search the current frame and the frames before and after the current frame, obtain the smallest spectral component of the previous frames and the smallest spectral component of the next frames, take the maximum value of these two values ​​as the current reference value, and calculate the speech existence probability P(n, m) based on the current reference value: Among them, T s (n, m) is the current reference value, μ is the fixed coefficient, and μ is the threshold coefficient; The noise spectrum d(n, m) is calculated based on P(n, m): Among them, ω is the compensation factor and γ is the constant coefficient.

7. A piano data evaluation system based on machine learning according to claim 6, characterized in that: Based on the noise spectrum, the posterior signal-to-noise ratio and the prior signal-to-noise ratio are obtained, and the gain factor is calculated based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio. The process includes: Obtain a noise power spectrum based on the noise spectrum, and obtain a piano audio power spectrum, and record the ratio of the piano audio power spectrum to the noise power spectrum as a priori signal-to-noise ratio; The ratio of the power spectrum of the piano audio after noise reduction to the power spectrum of the noise is recorded as the posterior signal-to-noise ratio; The gain factor G(n, m) is calculated based on the priori signal-to-noise ratio and the a posteriori signal-to-noise ratio: G(n,m)=X(n,m) / (1-H(n,m))*θ; Among them, X(n, m) is the prior signal-to-noise ratio, H(n, m) is the posterior signal-to-noise ratio, and θ is the gain coefficient.

8. A piano data evaluation system based on machine learning according to claim 7, characterized in that: The piano audio data enhancement according to the gain factor specifically includes the following process: multiplying the gain factor by the piano audio data to obtain enhanced audio data, and then performing inverse Fourier transformation on the enhanced audio data and then superimposing them.

9. The piano data evaluation system based on machine learning according to claim 1, characterized in that: Substituting the enhanced piano audio data into the neural network-based evaluation model to obtain the evaluation score specifically includes the following process: Feature extraction: converting piano audio data into feature representation suitable for neural network processing, the feature representation is a spectrogram; Data preparation: Divide the preprocessed piano audio data into training set, validation set and test set; Model design: Design a suitable network structure according to task requirements, including the number of layers, number of neurons, and activation function; Training and optimization: Use training data to train the model and adjust network parameters through the back-propagation algorithm to minimize the loss function; Evaluation and adjustment: Use the validation set to evaluate model performance and adjust the model structure and parameters based on the evaluation results; Get the evaluation score of the piano audio data based on the preset tags.

10. A piano data evaluation method based on machine learning, characterized in that: A piano data evaluation system based on machine learning applicable to any one of claims 1 to 9, the method comprising: Acquire piano audio data, and obtain actual load amplitude response, audio receiving frequency and audio evaluation quantity based on the piano audio data; The audio spectrum distortion is calculated based on the actual load amplitude response and the audio receiving frequency, the audio quality index is calculated based on the audio evaluation quantity, and the loading coefficient JZ is calculated based on the spectrum distortion and the audio quality index. Whether the piano audio data meets the evaluation requirements is determined by the loading coefficient JZ, and the piano audio data that meets the evaluation requirements is loaded into the evaluation module; Receiving piano audio data and conducting evaluation: preprocessing the piano audio data, and performing noise estimation on the preprocessed piano audio data to obtain a noise spectrum, obtaining a posterior signal-to-noise ratio and a priori signal-to-noise ratio based on the noise spectrum, and calculating a gain factor based on the posterior signal-to-noise ratio and the priori signal-to-noise ratio, and enhancing the piano audio data based on the gain factor, substituting the enhanced piano audio data into an evaluation model based on a neural network to obtain an evaluation score; Get evaluation scores and historical evaluation scores, and generate evaluation reports.

Images