Chinese language external education training system

Abstract

The invention discloses a Chinese language foreign education training system, which relates to the technical field of education training and comprises an audio set acquisition unit, an audio analysis unit and an audio comparison unit. The audio set acquisition unit is used for acquiring real-time practice audio of a user during Chinese language practice and preprocessing the real-time practice audio to obtain standard practice audio; the audio analysis unit is used for determining a preliminary comparison coefficient of the standard practice audio according to the coding result of each practice audio block in the standard practice audio; and the audio comparison unit is used for determining a practice result of the user according to the preliminary comparison coefficient of the standard practice audio and transmitting the practice result to the user terminal. Through processing the practice audio, the pronunciation practice result of the user can be obtained, guidance is provided for the user to correct pronunciation, and the practice efficiency of the user is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of education and training, and in particular to a Chinese language foreign education and training system. Background Art

[0002] In Chinese language teaching, pronunciation is the first step in language learning and the foundation for subsequent Chinese language learning. Therefore, it is necessary to monitor and correct students' pronunciation in real time during Chinese language teaching. However, traditional pronunciation evaluation methods often rely on manual judgment, which has problems such as strong subjectivity and low efficiency. In recent years, although there have been some automatic pronunciation evaluation systems based on computer technology, due to the complex syllable combinations of Chinese, the accuracy of their evaluation results is low and needs to be further improved. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a Chinese language foreign education and training system.

[0004] The technical solution of the present invention is: a Chinese language foreign education and training system includes an audio set acquisition unit, an audio analysis unit and an audio comparison unit;

[0005] The audio set acquisition unit is used to collect the real-time practice audio of the user when practicing Chinese language, and pre-process the real-time practice audio to obtain standard practice audio;

[0006] The audio analysis unit is used to determine the preliminary comparison coefficient of the standard practice audio according to the encoding results of each practice audio block in the standard practice audio;

[0007] The audio comparison unit is used to determine the user's practice results based on the preliminary comparison coefficient of the standard practice audio, and transmit the practice results to the user terminal.

[0008] Further, the audio analysis unit determines the preliminary comparison coefficient of the standard practice audio according to the encoding results of each practice audio block in the standard practice audio, including the following steps:

[0009] A1. Divide the standard practice audio into several practice audio blocks;

[0010] A2, generating a suffix audio set for each practice audio block;

[0011] A3, according to each practice audio block and the corresponding suffix audio set, each practice audio block is RS-encoded;

[0012] A4. Generate preliminary comparison coefficients of the standard practice audio according to the RS encoding results of each practice audio block.

[0013] The beneficial effect of the above further scheme is that in the present invention, the RS code of the audio, i.e., the Reed-Solomon code, mainly has a powerful error correction capability in the audio field. By processing the speech signal with a specific algorithm, a codeword containing redundant information (i.e., a suffix audio set) is generated. This redundant information enables the receiving end to recover the data according to the remaining data and the redundant information even if part of the data is damaged during the data transmission or storage process. Therefore, the present invention takes into account that the user's standard practice audio may still contain information such as noise, and uses the RS encoding result corresponding to the suffix audio set to generate a specific preliminary comparison coefficient for the standard practice audio as an important indicator for evaluating the audio similarity.

[0014] Furthermore, A2 includes the following sub-steps:

[0015] A21, extracting the spectrum centroid of each training audio block, and taking the mean of all spectrum centroids as the spectrum common value;

[0016] A22. Randomly select from the practice audio block frame speech signal, as a prefix signal set, where, Indicates the number of speech signal frames contained in the training audio block, Indicates rounding up;

[0017] A23, generating a signal feature matrix for the training audio block using the prefix signal set and the spectrum common value;

[0018] A24, extracting the L2 norm value of the signal feature matrix;

[0019] A25. Use the L2 norm value of the signal feature matrix to modify the prefix signal set to obtain the suffix signal set.

[0020] The beneficial effect of the above further scheme is that: in the present invention, the average point of the speech frame spectrum energy distribution of the training audio block is taken as the spectrum centroid. The mean of the spectrum centroids of all the training audio blocks is used to generate a specific signal feature matrix for the training audio block, and the norm eigenvalue of the matrix can be used to generate a parameter suffix signal set, so that the suffix signal set is more accurate and targeted.

[0021] Furthermore, in A23, the signal feature matrix The expression is:

[0022] ;

[0023] In the formula, represents the common value of the spectrum, represents the amplitude spectrum of the first frame of speech signal in the prefix signal set, Indicates the prefix signal set The amplitude spectrum of the frame speech signal, Indicates the prefix signal set The amplitude spectrum of the frame speech signal, The number of speech signal frames representing the prefix signal set, Represents an exponent.

[0024] Further, in A25, for the prefix signal set The calculation formula for correcting the frame speech signal is:

[0025] ;

[0026] In the formula, Indicates The linear prediction cepstral coefficients of the frame speech signal after correction, Indicates The linear prediction cepstral coefficients of the frame speech signal before correction, Represents the L2 norm value of the signal feature matrix.

[0027] Further, A4 includes the following sub-steps:

[0028] A41, according to the RS encoding results of all the practice audio blocks, obtain the encoding matrix of the standard practice audio;

[0029] A42. Use the values ​​of the coding matrix as preliminary comparison coefficients of the standard practice audio.

[0030] The basic form of RS code is usually To indicate that, Indicates the length of the codeword, that is, the number of symbols in each encoded data block, Indicates the number of symbols in the original data block.

[0031] Furthermore, in A41, the encoding matrix of the standard practice audio The expression is:

[0032] ;

[0033] ;

[0034] ;

[0035] In the formula, Indicates the codeword length of the RS encoding result corresponding to the first training audio block, Indicates the codeword length of the RS encoding result corresponding to the second training audio block, Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is represents the symbolic function, represents the element in the first row and second column of the encoding matrix, represents the element in the second row and first column of the encoding matrix, No. The number of symbols before RS encoding of the training audio block, represents the mean value function, Indicates rounding up.

[0036] Furthermore, the audio comparison unit determines the user's practice result according to the preliminary comparison coefficient of the standard practice audio, and transmits the practice result to the user terminal, including the following sub-steps:

[0037] B1. Calculate the Euclidean distance between the standard practice audio and the standard pronunciation audio, and process the Euclidean distance using the preliminary comparison coefficient of the standard practice audio;

[0038] B2. Determine whether the processed Euclidean distance is greater than or equal to a threshold value. If so, the user's practice result is qualified; otherwise, the user's practice result is unqualified.

[0039] This threshold can be set manually.

[0040] Furthermore, in B1, the calculation formula for processing the Euclidean distance is:

[0041] ;

[0042] In the formula, represents the processed Euclidean distance, Represents the Euclidean distance between the standard practice audio and the standard pronunciation audio. represents the preliminary comparison coefficient of the standard practice audio, Represents a constant.

[0043] The beneficial effects of the present invention are as follows: the present invention proposes a Chinese language foreign education and training system, which collects audio of users practicing pronunciation in the process of Chinese language learning, adds a suffix audio set to each audio block in the collected standard practice audio, and then encodes the audio block with the suffix audio set to obtain a preliminary comparison coefficient of the standard practice audio, which serves as an important indicator for comparing the standard practice audio with the standard pronunciation audio and is used to compare the two audios; the present invention can obtain the pronunciation practice result of the user by processing the practice audio, provide guidance for the user to correct pronunciation, and effectively improve the practice efficiency of the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a schematic diagram of the structure of the Chinese language foreign education and training system. DETAILED DESCRIPTION

[0045] The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0046] like Figure 1 As shown, the present invention provides a Chinese language foreign education and training system, including an audio set acquisition unit, an audio analysis unit and an audio comparison unit;

[0047] The audio set acquisition unit is used to collect the real-time practice audio of the user when practicing Chinese language, and pre-process the real-time practice audio to obtain standard practice audio;

[0048] The audio analysis unit is used to determine the preliminary comparison coefficient of the standard practice audio according to the encoding results of each practice audio block in the standard practice audio;

[0049] The audio comparison unit is used to determine the user's practice results based on the preliminary comparison coefficient of the standard practice audio, and transmit the practice results to the user terminal.

[0050] In an embodiment of the present invention, the audio analysis unit determines the preliminary comparison coefficient of the standard practice audio according to the encoding results of each practice audio block in the standard practice audio, including the following steps:

[0051] A1. Divide the standard practice audio into several practice audio blocks;

[0052] A2, generating a suffix audio set for each practice audio block;

[0053] A3, according to each practice audio block and the corresponding suffix audio set, each practice audio block is RS-encoded;

[0054] A4. Generate preliminary comparison coefficients of the standard practice audio according to the RS encoding results of each practice audio block.

[0055] In the present invention, the RS code of the audio, i.e., the Reed-Solomon code, mainly has a powerful error correction capability in the audio field. By processing the speech signal with a specific algorithm, a codeword containing redundant information (i.e., a suffix audio set) is generated. This redundant information enables the receiving end to recover the data according to the remaining data and the redundant information even if part of the data is damaged during the data transmission or storage process. Therefore, the present invention takes into account that the user's standard practice audio may still contain information such as noise, and uses the RS encoding result corresponding to the suffix audio set to generate a specific preliminary comparison coefficient for the standard practice audio as an important indicator for evaluating the audio similarity.

[0056] In this embodiment of the present invention, A2 includes the following sub-steps:

[0057] A21, extracting the spectrum centroid of each training audio block, and taking the mean of all spectrum centroids as the spectrum common value;

[0058] A22. Randomly select from the practice audio block frame speech signal, as a prefix signal set, where, Indicates the number of speech signal frames contained in the training audio block, Indicates rounding up;

[0059] A23, generating a signal feature matrix for the training audio block using the prefix signal set and the spectrum common value;

[0060] A24, extracting the L2 norm value of the signal feature matrix;

[0061] A25. Use the L2 norm value of the signal feature matrix to modify the prefix signal set to obtain the suffix signal set.

[0062] In the present invention, the average point of the speech frame spectrum energy distribution of the training audio block is taken as the spectrum centroid. The mean of the spectrum centroids of all the training audio blocks is used to generate a specific signal feature matrix for the training audio block, and the norm eigenvalue of the matrix can be used to parameterize the generation of the suffix signal set, making the suffix signal set more accurate and targeted.

[0063] In the embodiment of the present invention, in A23, the signal feature matrix The expression is:

[0064] ;

[0065] In the formula, represents the common value of the spectrum, represents the amplitude spectrum of the first frame of speech signal in the prefix signal set, Indicates the prefix signal set The amplitude spectrum of the frame speech signal, Indicates the prefix signal set The amplitude spectrum of the frame speech signal, The number of speech signal frames representing the prefix signal set, Represents an exponent.

[0066] In the embodiment of the present invention, in A25, the first The calculation formula for correcting the frame speech signal is:

[0067] ;

[0068] In the formula, Indicates The linear prediction cepstral coefficients of the frame speech signal after correction, Indicates The linear prediction cepstral coefficients of the frame speech signal before correction, Represents the L2 norm value of the signal feature matrix.

[0069] In this embodiment of the present invention, A4 includes the following sub-steps:

[0070] A41, according to the RS encoding results of all the practice audio blocks, obtain the encoding matrix of the standard practice audio;

[0071] A42. Use the values ​​of the coding matrix as preliminary comparison coefficients of the standard practice audio.

[0072] The basic form of RS code is usually To indicate that, Indicates the length of the codeword, that is, the number of symbols in each encoded data block, Indicates the number of symbols in the original data block.

[0073] In the embodiment of the present invention, in A41, the coding matrix of the standard practice audio The expression is:

[0074] ;

[0075] ;

[0076] ;

[0077] In the formula, Indicates the codeword length of the RS encoding result corresponding to the first training audio block, Indicates the codeword length of the RS encoding result corresponding to the second training audio block, Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is represents the symbolic function, represents the element in the first row and second column of the encoding matrix, represents the element in the second row and first column of the encoding matrix, No. The number of symbols before RS encoding of the training audio block, represents the mean value function, Indicates rounding up.

[0078] In the embodiment of the present invention, the audio comparison unit determines the user's practice result according to the preliminary comparison coefficient of the standard practice audio, and transmits the practice result to the user terminal, including the following sub-steps:

[0079] B1. Calculate the Euclidean distance between the standard practice audio and the standard pronunciation audio, and process the Euclidean distance using the preliminary comparison coefficient of the standard practice audio;

[0080] B2. Determine whether the processed Euclidean distance is greater than or equal to a threshold value. If so, the user's practice result is qualified; otherwise, the user's practice result is unqualified.

[0081] This threshold can be set manually.

[0082] In the embodiment of the present invention, in B1, the calculation formula for processing the Euclidean distance is:

[0083] ;

[0084] In the formula, represents the processed Euclidean distance, Represents the Euclidean distance between the standard practice audio and the standard pronunciation audio. represents the preliminary comparison coefficient of the standard practice audio, Represents a constant.

[0085] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A Chinese language foreign education and training system, characterized in that: It includes an audio set acquisition unit, an audio analysis unit and an audio comparison unit; The audio set acquisition unit is used to collect the real-time practice audio of the user when practicing Chinese language, and pre-process the real-time practice audio to obtain standard practice audio; The audio analysis unit is used to determine the preliminary comparison coefficient of the standard practice audio according to the encoding results of each practice audio block in the standard practice audio; The audio comparison unit is used to determine the user's practice result according to the preliminary comparison coefficient of the standard practice audio, and transmit the practice result to the user terminal.

2. The Chinese language foreign education and training system according to claim 1 is characterized in that: The audio analysis unit determines the preliminary comparison coefficient of the standard practice audio according to the encoding results of each practice audio block in the standard practice audio, including the following steps: A1. Divide the standard practice audio into several practice audio blocks; A2, generating a suffix audio set for each practice audio block; A3, according to each practice audio block and the corresponding suffix audio set, each practice audio block is RS-encoded; A4. Generate preliminary comparison coefficients of the standard practice audio according to the RS encoding results of each practice audio block.

3. The Chinese language foreign education and training system according to claim 2 is characterized in that: A2 includes the following sub-steps: A21, extracting the spectrum centroid of each training audio block, and taking the mean of all spectrum centroids as the spectrum common value; A22. Randomly select from the practice audio block frame speech signal, as a prefix signal set, where, Indicates the number of speech signal frames contained in the training audio block, Indicates rounding up; A23, generating a signal feature matrix for the training audio block using the prefix signal set and the spectrum common value; A24, extracting the L2 norm value of the signal feature matrix; A25. Use the L2 norm value of the signal feature matrix to modify the prefix signal set to obtain the suffix signal set.

4. The Chinese language foreign education and training system according to claim 3 is characterized in that: In A23, the signal characteristic matrix The expression is: ； In the formula, represents the common value of the spectrum, represents the amplitude spectrum of the first frame of speech signal in the prefix signal set, Indicates the prefix signal set The amplitude spectrum of the frame speech signal, Indicates the prefix signal set The amplitude spectrum of the frame speech signal, The number of speech signal frames representing the prefix signal set, Represents an exponent.

5. The Chinese language foreign education and training system according to claim 3 is characterized in that: In A25, the prefix signal is concentrated on the first The calculation formula for correcting the frame speech signal is: ； In the formula, Indicates The linear prediction cepstral coefficients of the frame speech signal after correction, Indicates The linear prediction cepstral coefficients of the frame speech signal before correction, Represents the L2 norm value of the signal feature matrix.

6. The Chinese language foreign education and training system according to claim 2, characterized in that: The A4 comprises the following sub-steps: A41, according to the RS encoding results of all the practice audio blocks, obtain the encoding matrix of the standard practice audio; A42. Use the values ​​of the coding matrix as preliminary comparison coefficients of the standard practice audio.

7. The Chinese language foreign education and training system according to claim 6, characterized in that: In the A41, the coding matrix of the standard practice audio The expression is: ； ； ； In the formula, Indicates the codeword length of the RS encoding result corresponding to the first training audio block, Indicates the codeword length of the RS encoding result corresponding to the second training audio block, Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is Indicates The codeword length of the RS encoding result corresponding to the training audio block is represents the symbolic function, represents the element in the first row and second column of the encoding matrix, represents the element in the second row and first column of the encoding matrix, No. The number of symbols before RS encoding of the training audio block, represents the mean value function, Indicates rounding up.

8. The Chinese language foreign education and training system according to claim 1 is characterized in that: The audio comparison unit determines the user's practice result according to the preliminary comparison coefficient of the standard practice audio, and transmits the practice result to the user terminal, including the following sub-steps: B1. Calculate the Euclidean distance between the standard practice audio and the standard pronunciation audio, and process the Euclidean distance using the preliminary comparison coefficient of the standard practice audio; B2. Determine whether the processed Euclidean distance is greater than or equal to a threshold value. If so, the user's practice result is qualified; otherwise, the user's practice result is unqualified.

9. The Chinese language foreign education and training system according to claim 8, characterized in that: In B1, the calculation formula for processing the Euclidean distance is: ； In the formula, represents the processed Euclidean distance, Represents the Euclidean distance between the standard practice audio and the standard pronunciation audio. represents the preliminary comparison coefficient of the standard practice audio, Represents a constant.

Images