CN101894560B - Reference source-free MP3 audio frequency definition objective evaluation method - Google Patents

Abstract

The invention relates to a reference source-free MP3 audio frequency definition objective evaluation method, which directly and objectively evaluates the definition quality of MP3 audio frequency. The method comprises the following steps of: firstly, acquiring an MP3 stereo audio frequency file and extracting a medium-high frequency index (MHFI), which affects the definition quality of the audio frequency, from the MP3 stereo audio frequency file, wherein the high audio frequency definition means that the audio frequency comprises more medium-high frequency components; secondly, computing possibilistic entropy of the medium-high frequency index of the audio frequency of each frame, wherein the entropy value reflects the rich degree of the audio frequency information, and the possibilistic entropy value of the highly definite audio frequency is higher than that of the indefinite audio frequency, so that the possibilistic entropy value can be used as a reference source-free audio frequency definition evaluation function; and finally, calculating and mapping the entropy value and defining the value of the audio frequency definition between 0 and 5. Experiments show that the method can effectively measure an objective evaluation value of the reference source-free MP3 stereo audio frequency definition, and the value is close to a subject evaluation value and is consistent with the hearing feeling of human ears.

Description

An objective evaluation method of MP3 audio clarity without reference source

technical field

The invention relates to the objective evaluation of audio quality, in particular to an objective evaluation method for the clarity of MP3 (MPEG1-layer3) audio without a reference source.

Background technique

An objective evaluation technology of audio clarity without a reference source in the present invention is to directly evaluate the clarity quality of an output signal of an audio system in the range of 0-5.

The objective evaluation of audio quality can be divided into evaluation based on input-output and evaluation based on output from the evaluation structure. Most of the work focuses on the evaluation based on input-output, which is to judge the quality of speech quality by the error between the input signal and output signal of the speech system, which is a kind of error measurement; while the evaluation based on output is only Quality evaluation is performed based on the output signal of the speech system, without reference sources; on the other hand, from the evaluation content, it can be divided into comprehensive evaluation of the overall audio and sub-indices (such as: clarity, fullness, brightness, softness, etc. ) for evaluation. The objective evaluation method of audio quality based on input-output is relatively perfect at present, and the evaluation based on output starts relatively late and basically evaluates the overall quality of audio in terms of evaluation content. There are mainly perceptual linear prediction methods, based on fuzzy multi-class support vectors The evaluation method of the machine, the method of measuring the density distribution characteristics of the spectrogram, etc. At present, there is no method for evaluating the quality of the sub-indices of the output signal. The present invention is just to study the objective evaluation of the clarity index of the MP3 audio signal without a reference source.

The objective evaluation method of audio clarity without a reference source proposed by the present invention improves the shortcomings of time-consuming, labor-intensive and high-cost subjective evaluation, and also solves the problem that the currently dominant objective evaluation method based on input-output is sometimes difficult to provide reference signals The unfavorable factors can further provide a reference for the objective evaluation of other indicators such as audio fullness and brightness. At the same time, various indicators can also be used as high-level perception parameters for speech recognition and classification retrieval in the MP3 compressed domain.

Contents of the invention

The purpose of the present invention is to provide an objective evaluation method of MP3 audio clarity without a reference source, providing a criterion for the best audio clarity quality. Directly extract the characteristic parameter reflecting the audio clarity from the MP3 compressed data - mid-high frequency index (MHFI), and calculate the possibility entropy value of MHFI, through the statistical mapping of the entropy value, the score of the audio clarity to be tested is limited to Between 0 and 5 points, to achieve an objective evaluation of the audio clarity index without a reference source.

The technical solution adopted by the present invention to solve the technical problem is: first extract the mid-high frequency index from the MP3 compressed audio data, calculate the possibility entropy value of the mid-high frequency index, and obtain the score of the audio clarity to be tested through statistical mapping.

The technical solution adopted by the present invention to solve the technical problem can be further improved. Firstly, the modified discrete cosine transform MDCT (Modified Discrete Cosine Transform) matrix is generated from the MP3 compressed audio data, and then the effective characteristic parameters are extracted from it: mid-high frequency index, and the possibility entropy value of the mid-high frequency index is calculated, and the entropy value is statistically mapped , to limit the score of the audio clarity to be tested between 0 and 5 points, so as to realize the objective evaluation of the audio clarity index without a reference source. The method specifically includes the following steps:

1) Preprocessing of MP3 compressed audio: including four parts: decoding frame header, side information reading, main data reading, Huffman decoding and quantization;

2) generate the MDCT matrix: find out the MDCT coefficients in each subband, arrange the coefficients in the subbands, and form three parts of the matrix;

3) Extraction of characteristic parameters in the compressed domain: MHFI (Medium-high frequency index);

4) Calculate the possibility entropy (E) of the medium and high frequency indices:

5) Statistical mapping: Statistically calculate the possibility entropy value of the medium and high frequency index, map it to the interval of 0-5 and output it.

The beneficial effects of the present invention are: directly extracting effective feature parameters from MP3 compressed audio data, compared with extracting features after decompressing the compressed data, the algorithm is simpler, and the calculation time is saved; the time-consuming, laborious and high cost of subjective evaluation is compensated At the same time, it also solves the unfavorable factors that the currently dominant objective evaluation method based on input-output is sometimes difficult to provide reference signals, which can further provide reference for the objective evaluation of other indicators such as audio fullness and brightness. At the same time, various The item index is used as a high-level perception parameter for speech recognition and retrieval in MP3 compression domain.

Description of drawings

Fig. 1 is a flowchart of an objective evaluation method for MP3 audio clarity without a reference source in the present invention.

Figure 2 is a comparison chart of subjective and objective evaluation scores.

Detailed ways

A preferred example of the present invention is described as follows in conjunction with accompanying drawing 1: this MP3 audio frequency clarity objective evaluation method without reference source, divides five steps altogether:

Step 1: MP3 compressed domain audio data processing

Compressed domain audio data processing can be divided into: frame header information reading, side information reading, main data reading, Huffman decoding and quantization.

1) Read frame header information

a) Define a structure for storing frame header information;

b) read the synchronization information in the frame;

c) synchronizing the decoder with the data stream;

d) Determine the starting position of the frame data and store the frame header information;

2) Reading of side information

a) Define the structure for storing side information;

b) The start position of the side information is determined by the end position of the frame header;

c) store side information;

3) Master data read

a) Define the structure for storing the scaling factor, and store the size of the main data;

b) Calculate the master data length;

c) Apply for the memory space of the main data length;

d) read master data;

e) read scaling factor;

4) Huffman decoding and dequantization

a) Define an array is[32][18] that stores Huffman decoded data in a particle;

b) Determine the starting position of the Huffman data in the main data according to the side information;

c) Decode the Huffman data and put the decoded data in is[32][18];

d) Dequantize the data in is[32][18] and store it in is[32][18].

Step 2: Generate MDCT coefficient matrix

The data of each particle consists of 32 subbands and each subband contains 18 coefficients. According to the principle of frequency distribution from low to high, each particle can form a 32×18 matrix. The process is as follows:

1. Find the coefficients of each subband

a) Find the coefficients S _i of the subbands in is[32][18], 32 in total;

b) Define the coefficients in the S _i subband as S _i [j], and each subband has 18 coefficients.

2. Form a row vector

a) Rearrange the coefficients in S _i according to the principle of frequency, and still store them in S _i [j];

b) S _i [j] after each sub-band arrangement is regarded as a row vector in the matrix.

3. Form a matrix

a) Combine the S _i [j] row vectors according to the sub-band numbers to form 32×18M[i][j];

b) According to the above principle, the MDCT coefficient matrix of two particles in one frame is expressed as M ₁ [i][j], M ₂ [i][j]

Step 3: Extraction of characteristic parameters in the compressed domain

The extracted compressed domain features are: Medium-high frequency index MHFI (Medium-high frequency index) parameters, the specific calculation steps are as follows:

a) Calculate the sum of the squares of each particle modified discrete cosine transform coefficient of the MP3 audio:

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 32</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ;</span>$

In the formula, i and j represent the serial number of the sideband and the serial number of the coefficient in the sideband; M[i][j] is the MDCT coefficient value.

b) Calculate the sum of squares of high-frequency MDCT coefficients in each particle:

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ;</span>$

Among them: the range of coefficient serial numbers can be moderately adjusted in a small range according to the selected medium and high frequency bands;

c) Define the medium-high frequency index MHFI (Medium-high frequency index) of each particle as:

$\mathrm{<span\; class="notranslate">\; MHFI</span>}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 32</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}$

Step 4: Calculate the possibility entropy E of the medium and high frequency indices

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$

Wherein N is the total frame number of the test audio, p _i is the mid-high frequency index value, since each frame contains two particles and the test audio is MP3 two-channel audio data, so each frame corresponds to four mid-high frequency index values;

Step Five: Statistical Mapping

a) Find the mean value EM of the high-frequency exponential possibility entropy in the entire audio:

$\mathrm{<span\; class="notranslate">\; EM</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$

In the formula, N is the total frame number of the audio, and p _i is the value of the middle and high frequency index, because each frame contains two particles and the test audio is MP3 stereo audio data, and each frame corresponds to the value of four middle and high frequency indexes;

b) Map the mean value EM of the possibility entropy of the mid-to-high frequency index to the 0-5 interval and output it;

First, the mean value of the possibility entropy of the medium and high frequency indices is moderately amplified to obtain the SII, and then the SII value is mapped to the 0-5 interval through a nonlinear mapping function to obtain the articulation index AI (Articulation Index), among which the commonly used mapping function Also includes: square function, logarithmic function, intercept function, window function, threshold function, multi-valued quantization function;

AI=10/π*arctan(SII)

Finally output the sharpness score.

Experimental results

The audio data used in this experiment is MP3 stereo data, and its sampling frequency is 44.1KHz. Divide the audio data into three groups, and each group consists of four pieces of audio data with the same content but the subjective perception of clarity and quality decreases successively. According to the above-mentioned objective evaluation method of audio clarity quality, the corresponding objective evaluation score of clarity can be obtained, and the data comparison result table with the subjective evaluation score is as follows:

Table 1: Comparison of subjective and objective evaluation scores of MP3 stereo audio clarity quality The comparison of subjective and objective evaluation scores is shown in Figure 2.

It can be clearly seen from the comparison chart of subjective and objective evaluation scores of audio clarity without a reference source that the objective evaluation method for MP3 audio clarity without a reference source of the present invention can effectively calculate the clarity score of the corresponding audio frequency for MP3 compressed data. value, and is very close to the subjective evaluation score, which is in line with the auditory experience of the human ear.

Claims (8)

1. a kind of MP3 audio clarity objective evaluation method without reference source, it is characterized in that: at first obtain modified discrete cosine transform coefficient by decoding MP3 compressed audio part, secondly to these data calculation frequency domain middle and high frequency index (MHFI), i.e. : the ratio of the energy of the high-frequency components in each frame of compressed-domain audio signal to the total energy of each frame in the frequency domain; then the possibility entropy function is selected as the evaluation function of the no-reference MP3 audio clarity, and finally the possibility entropy value of the mid-high frequency index is calculated Statistical mapping yields an objective evaluation score for clarity. 2. a kind of MP3 audio clarity objective evaluation method without reference source according to claim 1, is characterized in that: concrete operation steps are as follows: a) Preprocessing of MP3 compressed audio: decoding frame header, reading side information, reading main data, Huffman decoding and quantization; b) Generate a modified discrete cosine transform MDCT matrix: find out the modified discrete cosine transform coefficients in each subband, and arrange the coefficients in the subbands to form a matrix; c) Extraction of characteristic parameters in the compressed domain: medium and high frequency index MHFI, namely: the ratio of the energy of the high frequency components in the compressed domain audio signal of each frame to the total energy of the frequency domain of each frame: $\mathrm{<span\; class="notranslate">\; MHFI</span>}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 32</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}$ In the formula, i and j represent the serial number of the sideband and the serial number of the coefficient in the sideband, and the value of the sideband serial number i in the numerator can be moderately adjusted in a small range according to the selected medium and high frequency band; M[i][j] is the MDCT coefficient value ; d) Calculate the possibility entropy E of the medium and high frequency indices: $\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$ In the formula, k represents the number of different components, and p _i represents the possibility of occurrence of the i-th component, which differs from Shannon entropy in that the possibility entropy no longer needs to satisfy the constraint that the sum of the occurrence probabilities of each component must be 1; e) Statistical mapping: perform average statistics on the calculated possible entropy values of the medium and high frequency indices, and map them to the 0-5 interval; i. Find the mean value EM of the high-frequency exponential possibility entropy in the entire audio: $\mathrm{<span\; class="notranslate">\; EM</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ;</span>$ In the formula: N is the total frame number of the audio, p _i is the value of the mid-high frequency index, because each frame contains two particles and the test audio is MP3 stereo audio data, and each frame corresponds to four mid-high frequency indices; ii. Map the mean value EM of the possibility entropy of the medium and high frequency indices to the 0-5 interval; Firstly, the mean value of the possibility entropy of the medium and high frequency indices is moderately amplified to obtain the SII, and then the SII value is mapped to the range of 0-5 through a nonlinear mapping function to obtain the AI of the clarity index. The commonly used mapping functions also include: square function, logarithmic function, intercept function, window function, threshold function, multi-valued quantization function;

AI=10/π*arctan(SII) f) Output sharpness score: that is, the sharpness index AI value obtained by statistical mapping. 3. a kind of MP3 audio clarity objective evaluation method without reference source according to claim 2, is characterized in that: described step a) the specific implementation method of MP3 compressed audio preprocessing is: a) Read frame header information, b) reading of side information, c) master data read, d) Huffman decoding and dequantization. 4. a kind of MP3 audio clarity objective evaluation method without reference source according to claim 2, is characterized in that: described step b) generates MDCT matrix concrete implementation method is: 1), find out each sub-band coefficient; a) Find out the coefficients of the subbands in each granular Huffman decoding data array, a total of 32; b) Define the coefficients in the i-th subband as S _i [j], and each subband has 18 coefficients; 2), forming a row vector: a) Rearrange the coefficients in the i-th subband according to the principle of high and low frequencies, and still store them in S _i [j]; b) Treat S _i [j] after each sub-band is arranged as a row vector in the matrix; 3), forming a matrix a) Combine the S _i [j] row vectors according to the sub-band numbers to form a 32×18 M[i][j]; b) According to the above principles, the MDCT coefficient matrices of two particles in one frame are expressed as M ₁ [i][j], M ₂ [i][j]. 5. a kind of MP3 audio clarity objective evaluation method without reference source according to claim 2, is characterized in that: described step c) the concrete implementation method that compressed domain characteristic parameter extracts is: 1), MHFI (Medium-high frequency index) parameters a) Calculate the sum of the squares of each particle modified discrete cosine transform coefficient of the MP3 audio: $\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 32</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ;</span>$ In the formula, i and j represent the serial number of the sideband and the serial number of the coefficient in the sideband; M[i][j] is the MDCT coefficient value; b) Calculate the sum of squares of MDCT coefficients in the high-frequency band of each particle: $\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ;</span>$ Among them: the range of coefficient serial numbers can be moderately adjusted in a small range according to the selected medium and high frequency bands; c) Define the middle and high frequency index MHFI of each particle as: $\mathrm{<span\; class="notranslate">\; MHFI</span>}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 32</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 18</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}<span\; class="notranslate">\; .</span>$ 6. a kind of MP3 audio clarity objective evaluation method without reference source according to claim 2, is characterized in that: described step d) the concrete method of calculating the possibility entropy of middle and high frequency index is as follows: $\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$ Wherein N is the total number of frames of the test audio, and _pi is the mid-high frequency index. Since each frame contains two particles and the test audio is MP3 stereo audio data, each frame corresponds to four mid-high frequency index values. 7. a kind of MP3 audio clarity objective evaluation method without reference source according to claim 2, is characterized in that: the concrete realization method of described step e) statistical mapping is: a) Find the mean value EM of the high-frequency exponential possibility entropy in the entire audio: $\mathrm{<span\; class="notranslate">\; EM</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$ In the formula, N is the total frame number of the audio, and p _i is the value of the middle and high frequency index, because each frame contains two particles and the test audio is MP3 stereo audio data, and each frame corresponds to the value of four middle and high frequency indexes; b) Map the mean value EM of the possibility entropy of the middle and high frequency indices to the 0-5 interval; First, the mean value of the possibility entropy of the medium and high frequency indices is moderately amplified to obtain the SII value, and then the SII value is mapped to the 0-5 interval through a nonlinear mapping function to obtain the AI of the clarity index value. The commonly used mapping functions also include: Square function, logarithmic function, intercept function, window function, threshold function, multi-valued quantization function; AI=10/π*arctan(SII). 8. A kind of MP3 audio definition objective evaluation method without reference source according to claim 2, characterized in that: the definition score in the step f) is the definition index AI value that statistical mapping obtains.

Images