A speech recognition error correction method and system based on artificial intelligence

Abstract

The invention discloses a speech recognition error correction method and system based on artificial intelligence, which relates to the field of artificial intelligence technology, and comprises a control center, wherein the control center is connected with a speech acquisition module, a data processing module, an intelligent analysis module and a comprehensive monitoring module; the speech acquisition module acquires user speech data and sample speech data; in the data processing module, signal conversion and sorting are performed on the user speech data to obtain a user signal sequence, and a filtering wave coefficient is set to perform interval extraction on the user signal sequence to obtain a user extraction coefficient; in the intelligent analysis module, interval extraction, data stream conversion and verification transformation are performed on the sample speech data to obtain a monitoring sample matrix; in the comprehensive monitoring module, transformation and verification are performed on the user coefficient segment according to the monitoring sample matrix to obtain a verification user segment, and verification statistics and position correction are performed to obtain a corrected monitoring segment; the speech recognition accuracy and work efficiency are improved, and the user experience is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a speech recognition error correction method and system based on artificial intelligence. Background Art

[0002] Artificial intelligence refers to the theories, methods, technologies and application systems that use computer science and machine learning techniques to simulate, extend and expand human intelligence. Artificial intelligence systems can imitate human cognitive abilities such as learning, reasoning, recognition, understanding, communication and decision-making to perform various tasks.

[0003] Although current speech recognition technology has made significant progress, there are still some problems, such as recognition errors, accent differences, etc., which require error correction in the speech recognition process. Speech recognition error correction refers to the automatic editing and correction of the text output by the speech recognition system through artificial intelligence technology to improve recognition accuracy and reduce the error rate; therefore, using artificial intelligence to monitor errors in the speech recognition process has important theoretical and practical significance.

[0004] How to use artificial intelligence technology to perform signal conversion and sorting on the collected user voice data, obtain the user signal sequence, set the filter wave coefficient to perform interval extraction on the user signal sequence, obtain the user extraction coefficient and perform interval extraction, data stream conversion and verification transformation, and obtain the monitoring sample matrix; according to the monitoring sample matrix, transform and verify the user coefficient segment, obtain the verification user segment and perform verification statistics and position correction to obtain the corrected monitoring segment, is the problem we need to solve. To this end, a speech recognition error correction method and system based on artificial intelligence is now provided. Summary of the invention

[0005] The purpose of the present invention is to provide a speech recognition error correction method and system based on artificial intelligence.

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

[0007] A speech recognition error correction system based on artificial intelligence comprises a control center, wherein the control center is connected with a speech acquisition module, a data processing module, an intelligent analysis module and a comprehensive monitoring module;

[0008] The voice collection module is used to collect user voice data and sample voice data;

[0009] The process of the voice collection module collecting user voice data and sample voice data includes:

[0010] Set up a login terminal, register and log in the user through the login terminal, and mark the user who has completed the identity login as a collector;

[0011] Set up a sample collection end and an original collection end, collect the comprehensive information of the collector through the sample collection end, and obtain the user voice data;

[0012] Set a collection instruction, send the obtained collection instruction to the user, collect samples from the user through the original collection end, and obtain sample voice data.

[0013] The process of obtaining the user extraction coefficient includes:

[0014] Performing signal conversion on the obtained user voice data to obtain a user voice signal;

[0015] Setting a statistical period, sorting the obtained user voice signals based on the statistical period, and obtaining a user signal sequence;

[0016] Setting the filter coefficient according to the obtained statistical period and user signal sequence;

[0017] Setting a transfer parameter and a parameter interval, and obtaining a transfer number according to the obtained parameter interval, transfer parameter and user signal sequence;

[0018] The obtained filtering wave coefficient is adjusted and divided according to the obtained adjustment number to obtain the filtering interval coefficient;

[0019] The obtained filtering interval coefficient and the user signal sequence are subjected to interval extraction to obtain a user extraction coefficient.

[0020] The process of obtaining the user coefficient stream and the sample coefficient stream includes:

[0021] Performing signal conversion on the obtained sample speech data to obtain a sample speech signal;

[0022] Sorting the obtained sample speech signals based on the statistical period to obtain a sample signal sequence;

[0023] Performing interval extraction on the obtained filtering interval coefficient and the sample signal sequence to obtain a sample extraction coefficient;

[0024] The obtained user extraction coefficient and sample extraction coefficient are converted into data streams to obtain a user coefficient stream and a sample coefficient stream.

[0025] The process of obtaining the monitoring sample matrix includes:

[0026] Dividing the obtained user coefficient stream and sample coefficient stream into stream segments to obtain user coefficient segments and sample coefficient segments;

[0027] Perform flow segment marking on the obtained sample coefficient segment to obtain a characteristic sample segment;

[0028] Perform feature combination on the obtained feature sample segments to obtain a sample feature matrix;

[0029] The obtained sample feature matrix is ​​subjected to verification transformation to obtain a monitoring sample matrix.

[0030] The process of obtaining a revised monitoring segment includes:

[0031] Performing a transposition transformation on the obtained monitoring sample matrix to obtain a transposed monitoring matrix;

[0032] Multiplying the obtained user coefficient segment with the transposed monitoring matrix to obtain a verification user segment;

[0033] Perform verification statistics on the obtained verification user segments to obtain qualified verification segments and abnormal verification segments;

[0034] According to the obtained abnormal check segment, the user coefficient segment is compared with the error code to obtain the error code position;

[0035] The position of the user coefficient segment is corrected according to the obtained error code position to obtain a corrected monitoring segment.

[0036] Based on the above-mentioned speech recognition error correction system based on artificial intelligence, the present invention also provides a speech recognition error correction method based on artificial intelligence, comprising the following steps:

[0037] Step 1: Collect user voice data and sample voice data;

[0038] Step 2: Perform signal conversion and sorting on the user voice data to obtain a user signal sequence, set the filter wave coefficient and adjust and divide it to obtain the filter interval coefficient, perform interval extraction on the user signal sequence according to the filter interval coefficient, and obtain the user extraction coefficient;

[0039] Step 3: Perform signal conversion and interval extraction on the sample voice data to obtain sample extraction coefficients, perform data stream conversion on the user extraction coefficients and the sample extraction coefficients to obtain user coefficient streams and sample coefficient streams, obtain sample coefficient segments by dividing the sample coefficient streams into stream segments, perform stream segment marking and check transformation on the obtained sample coefficient segments, and obtain a monitoring sample matrix;

[0040] Step 4: Perform transformation verification on the user coefficient segment according to the monitoring sample matrix to obtain the verification user segment, perform verification statistics on the obtained verification user segment, obtain qualified verification segments and abnormal verification segments, perform error code comparison and position correction on the abnormal verification segment, and obtain the corrected monitoring segment.

[0041] Compared with the prior art, the invention has the following beneficial effects: performing signal conversion and sorting on the collected user voice data to obtain a user signal sequence, setting a filter wave coefficient to perform interval extraction on the user signal sequence to obtain a user extraction coefficient, performing signal conversion and interval extraction on the sample voice data to obtain a sample extraction coefficient; converting the collected user voice data and the sample voice data into the same data form, which is conducive to improving the standardization of the user voice data verification;

[0042] The sample extraction coefficients are subjected to data stream conversion and verification transformation to obtain a monitoring sample matrix; the user coefficient segments are transformed and verified according to the monitoring sample matrix to obtain a verified user segment, and verification statistics and position correction are performed to obtain a corrected monitoring segment; the obtained user voice data and sample voice data are converted into a matrix form that is convenient for error correction, which is beneficial to improving the accuracy of voice recognition, facilitating standardized error correction of user voice data, and improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than 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.

[0046] like Figure 1 As shown, a speech recognition and error correction system based on artificial intelligence includes a control center, and the control center is connected to a speech acquisition module, a data processing module, an intelligent analysis module and a comprehensive monitoring module;

[0047] The voice collection module is used to collect user voice data and sample voice data, and the specific process includes:

[0048] Set up a login terminal and register and log in users through the login terminal;

[0049] Enter the user's personal information and login password through the login terminal, and the control center will review the personal information, grant login instructions to the personal information that has passed the review, and generate a user account based on the personal information;

[0050] According to the obtained login instruction, the obtained user account and login password are uploaded to the login terminal for identity login, and the user who has completed the identity login is marked as the collector;

[0051] Set the sample collection end and the original collection end;

[0052] The comprehensive information of the collector is collected through the sample collection terminal to obtain the user's voice data;

[0053] Set a collection instruction, send the obtained collection instruction to the user, collect samples from the user through the original collection end, and obtain sample voice data;

[0054] It should be further explained that, in the specific implementation process, the collection instruction means that pronunciation is performed to the user by setting standard pronunciation materials, such as numbers, words, phrases and sentences, to ensure that the collected sample voice data is representative; the sample collection means that the user reads aloud according to the obtained collection instruction, and then the user's pronunciation data is collected through the original collection end to obtain the sample voice data.

[0055] The data processing module is used to perform signal conversion and sorting on the user voice data, obtain the user signal sequence, set the filter wave coefficient to perform interval extraction on the user signal sequence, and obtain the user extraction coefficient. The specific process includes:

[0056] Performing signal conversion on the obtained user voice data to obtain a user voice signal;

[0057] Setting a statistical period, sorting the obtained user voice signals based on the statistical period, and obtaining a user signal sequence;

[0058] Setting the filter coefficient according to the obtained statistical period and user signal sequence;

[0059] The expression form of the filtering wave coefficient is a function form;

[0060] Setting a manipulation parameter, wherein the manipulation parameter is a parameter used to control the scaling and translation transformation of the filtered wave coefficient in the time dimension and the frequency dimension;

[0061] Setting a parameter interval according to the obtained user voice signal, wherein the parameter interval represents the distance between two adjacent adjustment parameters;

[0062] The number of maneuvers is obtained according to the obtained parameter interval, maneuvering parameters and user signal sequence, and the obtained number of maneuvers is marked as DS, where: L represents the length of the user signal sequence, a represents the number of parameter intervals, JG represents the parameter interval, and Δb represents the mobilization parameter;

[0063] The obtained filtering wave coefficient is adjusted and divided according to the obtained adjustment number to obtain the filtering interval coefficient;

[0064] It should be further explained that, in the specific implementation process, the mobilization division means that the filter wave coefficients are equally divided according to the number of mobilizations obtained to obtain filter interval coefficients of equal length;

[0065] Performing interval extraction on the obtained filtering interval coefficient and the user signal sequence to obtain a user extraction coefficient;

[0066] It should be further explained that, in the specific implementation process, the interval extraction means uploading the obtained filter interval coefficients to the user signal sequence in sequence according to the order of mobilization and division, and convolving the filter interval coefficients with the corresponding positions of the user signal sequence to obtain the interval extraction coefficients, until all the filter interval coefficients are convolved with the user signal sequence, and convolving the obtained interval extraction coefficients to obtain the user extraction coefficients.

[0067] The intelligent analysis module is used to perform interval extraction, data stream conversion and verification transformation on the sample voice data to obtain a monitoring sample matrix. The specific process includes:

[0068] Performing signal conversion on the obtained sample speech data to obtain a sample speech signal;

[0069] Furthermore, the signal conversion rules for the sample voice data and the signal conversion rules for the user voice data are the same;

[0070] Sorting the obtained sample speech signals based on the statistical period to obtain a sample signal sequence;

[0071] Performing interval extraction on the obtained filtering interval coefficient and the sample signal sequence to obtain a sample extraction coefficient;

[0072] Performing data stream conversion on the obtained user extraction coefficient and sample extraction coefficient to obtain a user coefficient stream and a sample coefficient stream;

[0073] It should be further explained that, in the specific implementation process, the data stream conversion is to convert the obtained user extraction coefficient and sample extraction coefficient into a data stream composed of a number of binary code elements;

[0074] Dividing the obtained user coefficient stream and sample coefficient stream into stream segments to obtain user coefficient segments and sample coefficient segments;

[0075] Further, the stream segment division means dividing the binary code elements in the user coefficient stream into groups of 7 to obtain user coefficient segments, and if the last group does not meet 7, then complementing is performed forward until 7 are met; similarly, the binary code elements in the sample coefficient stream are divided into groups of 7 to obtain sample coefficient segments;

[0076] Perform flow segment marking on the obtained sample coefficient segment to obtain a characteristic sample segment;

[0077] Furthermore, the stream segment mark indicates that the 7 binary code elements of the obtained sample coefficient segment are grouped into the first 4 code elements as the front code elements, and the last 3 code elements as the feature code elements;

[0078] The obtained characteristic sample segments are marked as i=1, i=2, i=3, i=4, i=5, i=6, i=7, where i=1, i=2, i=3, i=4 are recorded as front-end code elements, and i=5, i=6, i=7 are recorded as characteristic code elements;

[0079] Perform feature combination on the obtained feature sample segments to obtain a sample feature matrix, and mark the obtained sample feature matrix as Z;

[0080] It should be further explained that, in the specific implementation process, the process of combining the features includes:

[0081] The obtained characteristic sample segments are grouped and combined into a group of three characteristic sample segments to obtain a sample characteristic matrix. The obtained sample characteristic matrix is ​​a 3×7 matrix, that is, a matrix with 3 rows and 7 columns;

[0082] The three rows of feature sample segments corresponding to the front-end codewords in the sample feature matrix are marked as the front-end matrix, that is, a 3×4 matrix, and the obtained front-end matrix is ​​marked as P;

[0083] The three rows of characteristic sample segments corresponding to the characteristic code elements in the sample characteristic matrix are marked as a characteristic matrix, that is, a 3×3 matrix, and the obtained characteristic matrix is ​​marked as I;

[0084] For example, the user feature matrix is but For the front-end matrix, is the feature matrix;

[0085] In particular, the corresponding sample feature matrix can be recorded as T = [P┆I]. If the last group of feature combinations of feature sample segments does not satisfy three feature sample segments, the sample feature matrix is ​​obtained by dividing the segments into groups of three from the end forward.

[0086] Perform a check transformation on the obtained sample feature matrix to obtain a monitoring sample matrix, and mark the obtained monitoring sample matrix as Z1 ;

[0087] It should be further explained that, in the specific implementation process, the verification transformation represents performing modulo 2 addition on each row in the obtained sample feature matrix until the feature matrix in the sample feature matrix after modulo 2 addition is a unit matrix, and then the sample feature matrix after modulo 2 addition is marked as a monitoring sample matrix.

[0088] The comprehensive monitoring module is used to perform transformation verification on the user coefficient segment according to the monitoring sample matrix, obtain the verified user segment and perform verification statistics and position correction to obtain the corrected monitoring segment. The specific process includes:

[0089] Performing transformation verification on the obtained user coefficient segment according to the obtained monitoring sample matrix to obtain a verification user segment;

[0090] It should be further explained that, in the specific implementation process, the transformation verification process includes:

[0091] The obtained user coefficient segment is marked as Y n , where Y n It consists of 7 binary symbols, marking the binary symbols in the user coefficient segment, i.e., j=1, j=2, j=3, j=4, j=5, j=6, j=7, and n represents the number of the user coefficient segment;

[0092] Performing a transposition transformation on the obtained monitoring sample matrix to obtain a transposed monitoring matrix;

[0093] The obtained user coefficient segment is multiplied by the transposed monitoring matrix to obtain the verification user segment, and the obtained verification user segment is marked as X, where:

[0094] Perform verification statistics on the obtained verification user segments to obtain qualified verification segments and abnormal verification segments;

[0095] Further, the verification statistics indicate that the verification user segment is divided into a qualified verification segment and an abnormal verification segment according to the status of the verification user segment;

[0096] When the verification user segment is a zero matrix, the verification user segment is a qualified verification segment, indicating that the user coefficient segment for transformation verification is correct and can be successfully compared with the sample voice data, that is, the recognized user's voice data is correct and has no errors;

[0097] When the verification user segment is a non-zero matrix, the verification user segment is an abnormal verification segment, indicating that the user coefficient segment for transformation verification is wrong, that is, the voice data of the recognized user is wrong;

[0098] Perform security error correction on the monitoring sample matrix according to the obtained abnormal check segment to obtain a corrected monitoring segment;

[0099] It should be further explained that, in the specific implementation process, the security error correction process includes:

[0100] After checking and counting the check user segments, the qualified check segments and abnormal check segments are obtained. The qualified check segments are X=000, and the abnormal check segments are X=001, X=010, X=100, X=011, X=101, X=110, and X=111.

[0101] According to the obtained abnormal check segment, the user coefficient segment is compared with the error code to obtain the error code position;

[0102] The process of obtaining the error code position includes:

[0103] When X=111, it means that the error position of the user coefficient segment is j=1;

[0104] When X=110, it means that the error position of the user coefficient segment is j=2;

[0105] When X=101, it means that the error position of the user coefficient segment is j=3;

[0106] When X=011, it means the error position of the user coefficient segment is j=4;

[0107] When X=100, it means that the error code position of the user coefficient segment is j=5;

[0108] When X=010, it means that the error position of the user coefficient segment is j=6;

[0109] When X=001, it means the error position of the user coefficient segment is j=7;

[0110] Performing position correction on the user coefficient segment according to the obtained error code position to obtain a corrected monitoring segment;

[0111] The position correction means changing the binary code element at the wrong position from the original 1 to 0 or from the original 0 to 1;

[0112] For example, when the user coefficient segment is 1010001, if the abnormal check segment obtained is X=010, that is, the error code position of the user coefficient segment is j=6, the binary code element at the j=6 position of the user coefficient segment is modified from 0 to 1, and 1010011 is obtained as the corrected monitoring segment;

[0113] Furthermore, the obtained user coefficient segments are checked and counted in sequence, and the positions of the obtained abnormal check segments are corrected to ensure that the obtained user voice data are correct.

[0114] Based on the above-mentioned speech recognition error correction system based on artificial intelligence, the present invention also provides a speech recognition error correction method based on artificial intelligence, comprising the following steps:

[0115] Step 1: Collect user voice data and sample voice data;

[0116] Step 2: Perform signal conversion and sorting on the user voice data to obtain a user signal sequence, set the filter wave coefficient and adjust and divide it to obtain the filter interval coefficient, perform interval extraction on the user signal sequence according to the filter interval coefficient, and obtain the user extraction coefficient;

[0117] Step 3: Perform signal conversion and interval extraction on the sample voice data to obtain sample extraction coefficients, perform data stream conversion on the user extraction coefficients and the sample extraction coefficients to obtain user coefficient streams and sample coefficient streams, obtain sample coefficient segments by dividing the sample coefficient streams into stream segments, perform stream segment marking and check transformation on the obtained sample coefficient segments, and obtain a monitoring sample matrix;

[0118] Step 4: Perform transformation verification on the user coefficient segment according to the monitoring sample matrix to obtain the verification user segment, perform verification statistics on the obtained verification user segment, obtain qualified verification segments and abnormal verification segments, perform error code comparison and position correction on the abnormal verification segment, and obtain the corrected monitoring segment.

[0119] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A speech recognition error correction system based on artificial intelligence, comprising a control center, characterized in that: The control center is connected to a voice acquisition module, a data processing module, an intelligent analysis module and a comprehensive monitoring module; The voice collection module is used to collect user voice data and sample voice data; The data processing module is used to perform signal conversion and sorting on the user voice data, obtain the user signal sequence, set the filter wave coefficient and adjust and divide it, obtain the filter interval coefficient, perform interval extraction on the user signal sequence according to the filter interval coefficient, and obtain the user extraction coefficient; The process of obtaining the user extraction coefficient includes: Performing signal conversion on the obtained user voice data to obtain a user voice signal; Setting a statistical period, sorting the obtained user voice signals based on the statistical period, and obtaining a user signal sequence; Setting the filter coefficient according to the obtained statistical period and user signal sequence; The expression form of the filtering wave coefficient is a function form; Setting a manipulation parameter, wherein the manipulation parameter is a parameter used to control the scaling and translation transformation of the filtered wave coefficient in the time dimension and the frequency dimension; Setting a parameter interval according to the obtained user voice signal, wherein the parameter interval represents the distance between two adjacent adjustment parameters; The number of maneuvers is obtained according to the obtained parameter interval, maneuvering parameters and user signal sequence, and the obtained number of maneuvers is marked as DS, where: , L represents the length of the user signal sequence, a represents the number of parameter intervals, JG represents the parameter interval, Indicates the mobilization parameters; The obtained filtering wave coefficients are divided according to the obtained number of divisions, wherein the division means that the filtering wave coefficients are divided equally according to the number of the obtained number of divisions to obtain filtering interval coefficients of equal length; The obtained filtering interval coefficient and the user signal sequence are subjected to interval extraction to obtain the user extraction coefficient: The interval extraction means uploading the obtained filter interval coefficients to the user signal sequence in sequence according to the order of the mobilization division, and convolving the filter interval coefficients with the corresponding positions of the user signal sequence to obtain the interval extraction coefficients, until all the filter interval coefficients are convolved with the user signal sequence, and convolving the obtained interval extraction coefficients to obtain the user extraction coefficients; The intelligent analysis module is used to perform signal conversion and interval extraction on the sample voice data, perform signal conversion on the obtained sample voice data to obtain a sample voice signal, perform signal conversion on the sample voice data according to the same conversion rule as that for the user voice data, sort the obtained sample voice signals based on the statistical period to obtain a sample signal sequence, perform interval extraction on the obtained filter interval coefficient and the sample signal sequence to obtain a sample extraction coefficient, perform data stream conversion on the user extraction coefficient and the sample extraction coefficient to obtain a user coefficient stream and a sample coefficient stream, obtain a sample coefficient segment by dividing the sample coefficient stream into stream segments, perform stream segment marking and check transformation on the obtained sample coefficient segment, and obtain a monitoring sample matrix; Dividing the obtained user coefficient stream and sample coefficient stream into stream segments to obtain user coefficient segments and sample coefficient segments; The comprehensive monitoring module is used to perform transformation verification on the user coefficient segment according to the monitoring sample matrix to obtain the verification user segment, perform verification statistics on the obtained verification user segment to obtain the qualified verification segment and the abnormal verification segment, perform error code comparison and position correction on the abnormal verification segment to obtain the corrected monitoring segment.

2. The artificial intelligence-based speech recognition error correction system according to claim 1, characterized in that: The process of obtaining the user coefficient stream and the sample coefficient stream includes: Performing signal conversion on the obtained sample speech data to obtain a sample speech signal; Sorting the obtained sample speech signals based on the statistical period to obtain a sample signal sequence; Performing interval extraction on the obtained filtering interval coefficient and the sample signal sequence to obtain a sample extraction coefficient; The obtained user extraction coefficient and sample extraction coefficient are converted into data streams to obtain a user coefficient stream and a sample coefficient stream.

3. The artificial intelligence-based speech recognition error correction system according to claim 2, characterized in that: The process of obtaining the monitoring sample matrix includes: Dividing the obtained user coefficient stream and sample coefficient stream into stream segments to obtain user coefficient segments and sample coefficient segments; Perform flow segment marking on the obtained sample coefficient segment to obtain a characteristic sample segment; Perform feature combination on the obtained feature sample segments to obtain a sample feature matrix; The obtained sample feature matrix is ​​subjected to verification transformation to obtain a monitoring sample matrix.

4. The speech recognition error correction system based on artificial intelligence according to claim 3 is characterized in that: The process of obtaining a revised monitoring segment includes: Performing a transposition transformation on the obtained monitoring sample matrix to obtain a transposed monitoring matrix; Multiplying the obtained user coefficient segment with the transposed monitoring matrix to obtain a verification user segment; Perform verification statistics on the obtained verification user segments to obtain qualified verification segments and abnormal verification segments; According to the obtained abnormal check segment, the user coefficient segment is compared with the error code to obtain the error code position; The position of the user coefficient segment is corrected according to the obtained error code position to obtain a corrected monitoring segment.

5. A speech recognition error correction method of a speech recognition error correction system based on artificial intelligence according to any one of claims 1 to 4, characterized in that: The following steps are involved: Step 1: Collect user voice data and sample voice data; Step 2: Perform signal conversion and sorting on the user voice data to obtain a user signal sequence, set the filter wave coefficient and adjust and divide it to obtain the filter interval coefficient, perform interval extraction on the user signal sequence according to the filter interval coefficient, and obtain the user extraction coefficient; Step 3: Perform signal conversion and interval extraction on the sample voice data to obtain sample extraction coefficients, perform data stream conversion on the user extraction coefficients and the sample extraction coefficients to obtain user coefficient streams and sample coefficient streams, obtain sample coefficient segments by dividing the sample coefficient streams into stream segments, perform stream segment marking and check transformation on the obtained sample coefficient segments, and obtain a monitoring sample matrix; Step 4: Perform transformation verification on the user coefficient segment according to the monitoring sample matrix to obtain the verification user segment, perform verification statistics on the obtained verification user segment, obtain qualified verification segments and abnormal verification segments, perform error code comparison and position correction on the abnormal verification segment, and obtain the corrected monitoring segment.

Images