CN106682161A - System for Japanese pronunciation correction - Google Patents

System for Japanese pronunciation correction Download PDF

Info

Publication number
CN106682161A
CN106682161A CN201611218260.1A CN201611218260A CN106682161A CN 106682161 A CN106682161 A CN 106682161A CN 201611218260 A CN201611218260 A CN 201611218260A CN 106682161 A CN106682161 A CN 106682161A
Authority
CN
China
Prior art keywords
signal
pronunciation
field
dimension table
japanese
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201611218260.1A
Other languages
Chinese (zh)
Inventor
郝杰
杨圆
崔慧玉
张可玲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beihua University
Original Assignee
Beihua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beihua University filed Critical Beihua University
Priority to CN201611218260.1A priority Critical patent/CN106682161A/en
Publication of CN106682161A publication Critical patent/CN106682161A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
    • G06F16/63—Querying
    • G06F16/632—Query formulation
    • G06F16/634—Query by example, e.g. query by humming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
    • G—PHYSICS
    • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
    • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
    • G09B19/00—Teaching not covered by other main groups of this subclass
    • G09B19/06—Foreign languages

Abstract

The invention relates to a system for Japanese pronunciation correction. The system comprises a voice collection module used to collect a user's Japanese pronunciation, an RAM storage device and an MRAM storage device which are connected to a single chip microcomputer in a wired manner and store standard Japanese pronunciation, an analysis module which is in wired connection with the single chip microcomputer and analyzes and compares the user's Japanese pronunciation and the standard Japanese pronunciation, a processing module which is in wired connection with the single chip microcomputer and marks and processes characters of the user's nonstandard pronunciation, a network connecting module which is in wired connection with the single chip microcomputer and conducts wireless network connection, as well as a cloud server which is in wireless connection with the network connecting module via a GPRS wireless network and stores the user's wrong Japanese pronunciation. According to the invention, each Japanese character pronunciation spoken by the user can be analyzed and processed; in combination with the existing standard pronunciation; the user's pronunciation can be corrected, so a Japanese pronunciation study level of the user can be increased; and a pronunciation error database of the user can be established by the cloud server.

Description

A kind of Japanese pronunciation correction system
Technical field
The invention belongs to language learning technical field, more particularly to a kind of Japanese pronunciation correction system.
Background technology
With the development and the progress in epoch of society, increasing people learns Japanese, but existing learned lesson one As be entity course, its network courses is less, typically carries out imparting knowledge to students by teacher and spoken training exchange, utilizes interconnect after class Net carries out the study of Japanese pronunciation, but existing Japanese articulatory system can not determine different people according to the pronunciation of different user Between difference, it is impossible to the pronunciation difference of definite each character of understanding so that user does not know which character is sent out Sound is nonstandard, causes the word pronunciation learning quality of Japanese spoken not good.
The content of the invention
The present invention can not be analyzed process for solution is existing to each word pronunciation character, it is impossible to set up pronunciation error number A kind of Japanese pronunciation correction system is provided according to the technical problem in storehouse.
The present invention is adopted the technical scheme that to solve technical problem present in known technology：A kind of Japanese pronunciation is rectified Positive system, the Japanese pronunciation correction system includes：
For the voice acquisition module to gathering the pronunciation of user's Japanese；The voice acquisition module awareness apparatus are in independence Sampling period in echo signal x (t) is acquired, and digital quantization is carried out to signal with A/D modes；Then, to quantifying Signal x (i) afterwards carries out dimensionality reduction；Finally, the signal after dimensionality reduction is reconstructed；Wherein t is sampling instant, and i is after quantization Signal sorts；Dimensionality reduction is carried out to the signal after quantization, specifically to the signal after quantization by finite impulse response filter Difference equation Wherein h (0) ..., h (L-1) are filter coefficient, and design is based on The compressed sensing signals collecting framework of filtering, constructs following Teoplitz calculation matrix：Then observeIts Middle b1,…,bLRegard filter coefficient as；Submatrix ΦFTSingular value be gram matrix G (ΦF, T) and=Φ 'FTΦFTEigenvalue Arithmetic root, verify G (Φ F, T) all eigenvalue λ i ∈ (1- δK,1+δK), i=1 ..., T, then ΦFMeet RIP, and pass through SolveOptimization problem reconstructs original signal reconstructing original signal by linear programming method, that is, BP algorithm；For actual compression signal, the such as collection of voice signal, then Φ is changedFFor following form：
If signal has sparse on conversion basic matrix Ψ Property, then by solving such as following formula optimization problem, Accurate Reconstruction goes out original signal：Its Middle Φ is uncorrelated to Ψ, and Ξ is referred to as CS matrixes；
With voice acquisition module by input module wired connection, for being converted into the signal of the signal of telecommunication to the voice for gathering Modular converter；
With signal conversion module wired connection, for being analyzed the single-chip microcomputer of process to the signal of telecommunication；The single-chip microcomputer is used Process is amplified in the signal for obtaining signal conversion module transmission and to signal；Extract from every segment signal average, variance, The basic time domain parameter of accumulated value and peak value 4 of signal, determines whether to doubt by the difference of 4 parameter values of adjacent segment signal The ground floor decision-making that non-type situation of seemingly pronouncing occurs judges：The no person of wavelet packet denoising is down performed if having, execution is jumped to Obtain the signal of signal conversion module transmission；Wavelet packet denoising is carried out to the signal for gathering using Wavelet Packet Algorithm is improved；Utilize Improving Wavelet Packet Algorithm carries out WAVELET PACKET DECOMPOSITION and reconstruct to the signal for gathering, and obtains list band reconstruction signal；From the list of reconstruct Extract in subband signal：Time domain energy, time domain peak, frequency domain energy, frequency domain peak value, coefficient of kurtosis, variance, frequency spectrum and deflection system The parameter of 8 expression signal characteristics of number；Using principal component analytical method, select from above-mentioned parameter 3 to 8 can obvious expression sound The parameter composition characteristic vector of transmission signal feature, and these characteristic vectors are input to into support vector machine carry out decision-making judgement, That is second layer decision-making judgement, determines whether that non-type situation occurs according to the output of support vector machine；
With single-chip microcomputer wired connection, for storing the RAM memory and mram memory of Japanese RP；
With single-chip microcomputer wired connection, the enquiry module inquired about for the Japanese of user to be pronounced；
With single-chip microcomputer wired connection, be analyzed dividing for contrast with the Japanese pronunciation of standard for the Japanese of user to be pronounced Analysis module；The analysis module is used for construction multilevel iudge matrix two-by-two, according to 1~9 scaling law successively to each key element two-by-two Between carry out importance degree assignment, Judgement Matricies U=(uij)n×n, wherein uijExpression factor uiAnd ujSend out relative to standard The importance value of message breath, matrix U has property：uii=1, uij=1/uji, i, j=1,2 ..., n draw judgment matrix；By square Battle array X1~X5By row normalization, i.e.,：Calculating matrix Y is： Y matrix by rowss are added, by formulaDraw：W1=(2.652 0.686 0.253 0.409)T, W2=(1 1)T, W3=(1.273 0.371 0.221 2.135)T, W4=(1.9 0.319 0.781)T, W5=(2.121 0.604 0.275)T；Obtain and vector is normalized, by formulaObtain weight vector： Calculate the Maximum characteristic root λ of judgment matrixmax, by formula Calculate: According to public affairs FormulaConsistency check is carried out, is obtained：CI1=0.019, CI2=0, CI3=0.031, CI4=0.020, CI5=0.048；By formulaObtain CR1=0.022, CR2=0, CR3=0.035, CR4= 0.038, CR5=0.092；CR ＜ 0.1, are satisfied by coherence request, therefore the relative weighting of each factor Relative analyses go out the Japanese pronunciation of user and the Japanese of standard pronounces after consistency check Contrast；
With single-chip microcomputer wired connection, for the processing module that the character of the nonstandard pronunciation of user is marked and is processed； The processing module is used to receive safeguards more new command；According to the maintenance more new command obtain the nonstandard pronunciation information of user with And need to safeguard the dimension table information for updating；According to the dimension table configuration information that the dimension table acquisition of information pre-sets；Its In, need to safeguard source database, the dimension table that the dimension table for updating is located with described in the dimension table configuration information Need purpose data base and the dimension table operating right information of synchronization；According to the nonstandard pronunciation information of the user and described Dimension table operating right information, judges whether the nonstandard pronunciation information of the user meets the dimension table operating right information； If the nonstandard pronunciation information of the user meets the dimension table operating right information, need to safeguard the dimension table for updating to described It is updated operation；The dimension table being updated after operation is synchronized in the RAM memory and mram memory；
With single-chip microcomputer by data update module wired connection, for being updated the external memory of Japanese pronunciation；
With single-chip microcomputer wired connection, for carrying out the network connecting module of wireless network connection；
Wirelessly it is connected by GPRS wireless networks with network connecting module, the Japanese for user error to be pronounced is deposited The Cloud Server of storage.
Further, wavelet packet denoising and WAVELET PACKET DECOMPOSITION include with reconstruct：
Signals extension, to each layer signal of WAVELET PACKET DECOMPOSITION horizontal parabola continuation is entered；
If signal data is x (a), x (a+1), x (a+2), then the expression formula of continuation operator E is：
Eliminate list band un-necessary frequency composition；
By the signal after continuation and decomposition low pass filter h0Convolution, obtains low frequency coefficient, then calculates through HF-cut-IF Subprocessing, removes unnecessary frequency content, then carries out down-sampling, obtains next layer of low frequency coefficient；By the signal after continuation with Decompose high pass filter g0Convolution, obtains high frequency coefficient, then through LF-cut-IF operators process, remove unnecessary frequency into Point, then down-sampling is carried out, and next layer of high frequency coefficient is obtained, shown in HF-cut-IF operators such as formula (2), LF-cut-IF operators such as formula (3) shown in；
In (2), (3) formula, x (n) is 2jThe coefficient of wavelet packet, N on yardstickjRepresent 2jThe length of data on yardstick,K=0,1 ..., Nj-1；N=0,1 ..., Nj-1；
It is described to obtain list band reconstruction signal, specifically include：The high and low frequency coefficient for obtaining is up-sampled, Ran Houfen Not with high pass reconstruction filter g1With low-pass reconstruction filter h1Convolution, the signal for obtaining is used respectively HF-cut-IF, LF-cut- The process of IF operators, obtains list band reconstruction signal.
Further, in the processing module, dimension table operating right information includes：User with dimension table operating right Nonstandard pronunciation information mark；
It is described to judge whether the nonstandard pronunciation information of the user meets the dimension table operating right information, including：
Judge the nonstandard pronunciation information of the user whether in nonstandard of the user with dimension table operating right In sound message identification；
The maintenance more new command is for increase content instruction, change content instruction or deletes content instruction；
Before operation is updated to the dimension table for needing maintenance to update, including：
More new command is safeguarded according to described, it is determined that needs safeguard the field for updating, and the needs maintenance renewal is got Field identification；
The field configuration information for pre-setting is got according to the field identification and the dimension table configuration information；Its In, the field configuration information includes field contents ordering rule, field ordering information, the field restrictive condition of the field；
If the maintenance more new command is to increase content instruction, described to need to safeguard that the dimension table for updating is carried out more to described New operation, including：
Obtain the corresponding batch data content of the increase content instruction；
According to the batch data content, in one or more fields in the dimension table field contents are increased；
According to the field contents ordering rule, the field contents are ranked up；
According to the field ordering information, each field in dimension table is ranked up；
If the maintenance more new command is change content instruction, described to need to safeguard that the dimension table for updating is carried out more to described New operation, including：
Obtain the corresponding batch data content of the change content instruction；
According to the batch data content, in one or more fields in the dimension table field contents are changed；
If the maintenance more new command is to delete content instruction, described to need to safeguard that the dimension table for updating is carried out more to described New operation, including：
Field contents are deleted in one or more fields in the dimension table；
Also include：
Judge whether each field after the increase field contents, change field contents or deletion field contents meets The field restrictive condition；
If there is field to be unsatisfactory for the field restrictive condition, information is generated；The information is used to point out to be discontented with The Field Count of the foot field restrictive condition, and prompting is unsatisfactory for the field relevant information of the field restrictive condition；The word Section relevant information includes the field identification or field name of the field.
Further, the single-chip microcomputer and power module wired connection, for providing power supply.
Further, the single-chip microcomputer passes through output module and speaker wired connection, for the output of standard Japanese pronunciation.
Further, the single-chip microcomputer passes through output module and display screen wired connection, for the Japanese after labelling and process Shown.
The present invention has the advantages and positive effects that：The Japanese pronunciation correction system can analyze and process each from The Japanese character pronunciation that user sends, and with reference to existing RP, the pronunciation to user is corrected, and improves the day of user Language word pronunciation learning level, and the pronunciation error database of user is set up by Cloud Server.
The signal acquisition method of voice acquisition module of the present invention enhances the scope to signal collection；Improve data protection Safety and the accuracy of data signal；The information processing of processing module of the present invention, it is ensured that to nonstandard pronunciation information and When process, further increase the convenience and directiveness for using.The signal processing of single-chip microcomputer of the present invention be effectively guaranteed and When capture effective information and in time adjustment do not mark pronunciation information.The contrast judgment matrix that the present invention passes through analysis module, alignment True pronunciation has directive significance.
Description of the drawings
Fig. 1 is the theory diagram of Japanese pronunciation correction system provided in an embodiment of the present invention；
In figure：1st, voice acquisition module；2nd, input module；3rd, signal conversion module；4th, single-chip microcomputer；5th, RAM memory；6、 Mram memory；7th, enquiry module；8th, analysis module；9th, processing module；10th, data update module；11st, external memory；12nd, net Network link block；13rd, Cloud Server；14th, power module；15th, output module.
Specific embodiment
For the content of the invention, feature and effect of the present invention can be further appreciated that, following examples are hereby enumerated, and coordinate accompanying drawing Describe in detail as follows.
The structure of the present invention is explained in detail with reference to Fig. 1.
The Japanese pronunciation correction system includes：
For the voice acquisition module 1 to gathering the pronunciation of user's Japanese；The voice acquisition module awareness apparatus are only Echo signal x (t) is acquired in the vertical sampling period, and digital quantization is carried out to signal with A/D modes；Then, to amount Signal x (i) after change carries out dimensionality reduction；Finally, the signal after dimensionality reduction is reconstructed；Wherein t is sampling instant, after i is to quantify Signal sequence；Dimensionality reduction is carried out to the signal after quantization, specifically finite impulse response filter is passed through to the signal after quantization Difference equation Wherein h (0) ..., h (L-1) are filter coefficient, design base In the compressed sensing signals collecting framework of filtering, following Teoplitz calculation matrix is constructed：Then observeIts Middle b1,…,bLRegard filter coefficient as；Submatrix ΦFTSingular value be gram matrix G (ΦF, T) and=Φ 'FTΦFTEigenvalue Arithmetic root, verify G (Φ F, T) all eigenvalue λ i ∈ (1- δK,1+δK), i=1 ..., T, then ΦFMeet RIP, and pass through SolveOptimization problem reconstructs original signal reconstructing original signal by linear programming method, that is, BP algorithm；For actual compression signal, the such as collection of voice signal, then Φ is changedFFor following form：
If signal has sparse on conversion basic matrix Ψ Property, then by solving such as following formula optimization problem, Accurate Reconstruction goes out original signal：Its Middle Φ is uncorrelated to Ψ, and Ξ is referred to as CS matrixes.
With voice acquisition module 1 by the wired connection of input module 2, for being converted into the letter of the signal of telecommunication to the voice for gathering Number modular converter 3；
With the wired connection of signal conversion module 3, for being analyzed the single-chip microcomputer 4 of process to the signal of telecommunication；The single-chip microcomputer For obtaining the signal of signal conversion module transmission and being amplified to signal process；Average, side are extracted from every segment signal Difference, the accumulated value of signal and the basic time domain parameter of peak value 4, are judged whether by the difference of 4 parameter values of adjacent segment signal There is the ground floor decision-making that the non-type situation of doubtful pronunciation occurs to judge：The no person of wavelet packet denoising is down performed if having, is jumped to Perform the signal for obtaining signal conversion module transmission；Wavelet packet denoising is carried out to the signal for gathering using Wavelet Packet Algorithm is improved； WAVELET PACKET DECOMPOSITION and reconstruct are carried out to the signal for gathering using Wavelet Packet Algorithm is improved, list band reconstruction signal is obtained；From reconstruct List band signal in extract：Time domain energy, time domain peak, frequency domain energy, frequency domain peak value, coefficient of kurtosis, variance, frequency spectrum and partially The parameters of 8 expression signal characteristics of oblique coefficient；Using principal component analytical method, the obvious table of 3 to 8 energy of selection from above-mentioned parameter Show the parameter composition characteristic vector of sound emission signal characteristic, and these characteristic vectors are input to into support vector machine carry out decision-making and sentence Disconnected, i.e. second layer decision-making judges, determines whether that non-type situation occurs according to the output of support vector machine；
Wavelet packet denoising and WAVELET PACKET DECOMPOSITION include with reconstruct：
Signals extension, to each layer signal of WAVELET PACKET DECOMPOSITION horizontal parabola continuation is entered；
If signal data is x (a), x (a+1), x (a+2), then the expression formula of continuation operator E is：
Eliminate list band un-necessary frequency composition；
By the signal after continuation and decomposition low pass filter h0Convolution, obtains low frequency coefficient, then calculates through HF-cut-IF Subprocessing, removes unnecessary frequency content, then carries out down-sampling, obtains next layer of low frequency coefficient；By the signal after continuation with Decompose high pass filter g0Convolution, obtains high frequency coefficient, then through LF-cut-IF operators process, remove unnecessary frequency into Point, then down-sampling is carried out, and next layer of high frequency coefficient is obtained, shown in HF-cut-IF operators such as formula (2), LF-cut-IF operators such as formula (3) shown in；
In (2), (3) formula, x (n) is 2jThe coefficient of wavelet packet, N on yardstickjRepresent 2jThe length of data on yardstick,K=0,1 ..., Nj-1；N=0,1 ..., Nj-1；
It is described to obtain list band reconstruction signal, specifically include：The high and low frequency coefficient for obtaining is up-sampled, Ran Houfen Not with high pass reconstruction filter g1With low-pass reconstruction filter h1Convolution, the signal for obtaining is used respectively HF-cut-IF, LF-cut- The process of IF operators, obtains list band reconstruction signal.
With the wired connection of single-chip microcomputer 4, for storing the RAM memory 5 and mram memory 6 of Japanese RP；
With the wired connection of single-chip microcomputer 4, the enquiry module 7 inquired about for the Japanese of user to be pronounced；
With the wired connection of single-chip microcomputer 4, be analyzed contrast with the Japanese pronunciation of standard for the Japanese of user to be pronounced Analysis module 8；The analysis module is used for construction multilevel iudge matrix two-by-two, according to 1~9 scaling law successively to each key element two Importance degree assignment, Judgement Matricies U=(u are carried out between twoij)n×n, wherein uijExpression factor uiAnd ujRelative to standard The importance value of pronunciation information, matrix U has property：Draw judgment matrix；By matrix X1~X5By row normalization, i.e.,：Calculating matrix Y is： Y matrix by rowss are added, by formulaDraw：W1=(2.652 0.686 0.253 0.409)T, W2=(1 1)T, W3=(1.273 0.371 0.221 2.135)T, W4=(1.9 0.319 0.781)T, W5=(2.121 0.604 0.275)T；Obtain and vector is normalized, by formulaObtain weight vector： Calculate the Maximum characteristic root λ of judgment matrixmax, by formula Calculate: According to FormulaConsistency check is carried out, is obtained：CI1=0.019, CI2=0, CI3= 0.031, CI4=0.020, CI5=0.048；By formulaObtain CR1=0.022, CR2=0, CR3=0.035, CR4= 0.038, CR5=0.092；CR ＜ 0.1, are satisfied by coherence request, therefore the relative weighting of each factor Relative analyses go out the Japanese pronunciation of user and the Japanese of standard pronounces after consistency check Contrast；
With the wired connection of single-chip microcomputer 4, for the processing module that the character of the nonstandard pronunciation of user is marked and is processed 9；The processing module is used to receive safeguards more new command；The nonstandard pronunciation information of user is obtained according to the maintenance more new command And need to safeguard the dimension table information for updating；According to the dimension table configuration information that the dimension table acquisition of information pre-sets； Wherein, need to safeguard source database, the dimension that the dimension table for updating is located with described in the dimension table configuration information Table needs the purpose data base of synchronization and dimension table operating right information；According to the nonstandard pronunciation information of the user and institute Dimension table operating right information is stated, judges whether the nonstandard pronunciation information of the user meets the dimension table operating right letter Breath；If the nonstandard pronunciation information of the user meets the dimension table operating right information, need to safeguard the dimension for updating to described Degree table is updated operation；The dimension table being updated after operation is synchronized in the RAM memory and mram memory；
Dimension table operating right information includes：The nonstandard pronunciation information mark of user with dimension table operating right；
It is described to judge whether the nonstandard pronunciation information of the user meets the dimension table operating right information, including：
Judge the nonstandard pronunciation information of the user whether in nonstandard of the user with dimension table operating right In sound message identification；
The maintenance more new command is for increase content instruction, change content instruction or deletes content instruction；
Before operation is updated to the dimension table for needing maintenance to update, including：
More new command is safeguarded according to described, it is determined that needs safeguard the field for updating, and the needs maintenance renewal is got Field identification；
The field configuration information for pre-setting is got according to the field identification and the dimension table configuration information；Its In, the field configuration information includes field contents ordering rule, field ordering information, the field restrictive condition of the field；
If the maintenance more new command is to increase content instruction, described to need to safeguard that the dimension table for updating is carried out more to described New operation, including：
Obtain the corresponding batch data content of the increase content instruction；
According to the batch data content, in one or more fields in the dimension table field contents are increased；
According to the field contents ordering rule, the field contents are ranked up；
According to the field ordering information, each field in dimension table is ranked up；
If the maintenance more new command is change content instruction, described to need to safeguard that the dimension table for updating is carried out more to described New operation, including：
Obtain the corresponding batch data content of the change content instruction；
According to the batch data content, in one or more fields in the dimension table field contents are changed；
If the maintenance more new command is to delete content instruction, described to need to safeguard that the dimension table for updating is carried out more to described New operation, including：
Field contents are deleted in one or more fields in the dimension table；
Also include：
Judge whether each field after the increase field contents, change field contents or deletion field contents meets The field restrictive condition；
If there is field to be unsatisfactory for the field restrictive condition, information is generated；The information is used to point out to be discontented with The Field Count of the foot field restrictive condition, and prompting is unsatisfactory for the field relevant information of the field restrictive condition；The word Section relevant information includes the field identification or field name of the field.
With single-chip microcomputer 4 by the wired connection of data update module 10, for being updated the external memory 11 of Japanese pronunciation；
With the wired connection of single-chip microcomputer 4, for carrying out the network connecting module 12 of wireless network connection；
Wirelessly it is connected by GPRS wireless networks with network connecting module 12, the Japanese for user error to be pronounced is carried out The Cloud Server 13 of storage.
Further, the single-chip microcomputer 4 and the wired connection of power module 14, for providing power supply.
Further, the single-chip microcomputer 4 passes through output module 15 and speaker wired connection, for the defeated of standard Japanese pronunciation Go out.
Further, the single-chip microcomputer 4 passes through output module 15 and display screen wired connection, for the day after labelling and process Language is shown.
The structure of the present invention is further described with reference to operation principle.
The Japanese sent from user is gathered by voice acquisition module 1 to pronounce, input module 2 is sharp by the pronunciation for collecting The signal of telecommunication is converted into signal conversion module 3, in being subsequently sent to single-chip microcomputer 4, using enquiry module 7 by the day of required study Language finds out in RAM memory 5 and mram memory 6, is broadcast by speaker and display screen using output module 15 Put and show, analysis module 8 can analyze and process each and pronounce from the Japanese character that user sends, and with reference to existing mark Quasi- pronunciation, the pronunciation to user is corrected, and improves the Japanese word pronunciation learning level of user, and sets up use by Cloud Server 13 The pronunciation error database at family is labeled and is processed using 9 pairs of pronunciation characters not marked of processing module, and by display screen Show so that user is learnt, power module 14 provides power supply, using external memory 11 data update module 10 is passed through The pronunciation data of Japanese is updated.
The above is only, to presently preferred embodiments of the present invention, any pro forma restriction not to be made to the present invention, Every technical spirit according to the present invention is belonged to any simple modification made for any of the above embodiments, equivalent variations and modification In the range of technical solution of the present invention.

Claims (6)

1. a kind of Japanese pronunciation correction system, it is characterised in that the Japanese pronunciation correction system includes：
For the voice acquisition module to gathering the pronunciation of user's Japanese；The voice acquisition module awareness apparatus are adopted in independent Echo signal x (t) is acquired in the sample cycle, and digital quantization is carried out to signal with A/D modes；Then, after to quantization Signal x (i) carries out dimensionality reduction；Finally, the signal after dimensionality reduction is reconstructed；Wherein t is sampling instant, and i is the signal after quantifying Sequence；Dimensionality reduction, the difference for specifically finite impulse response filter being passed through to the signal after quantization are carried out to the signal after quantization EquationI=1 ..., M, wherein h (0) ..., h (L-1) is filter coefficient, and design is based on filtering Compressed sensing signals collecting framework, constructs following Teoplitz calculation matrix：Then observeI=1 ..., M, wherein b1,…,bLRegard filter coefficient as；Submatrix ΦFTSingular value be gram matrix G (ΦF, T) and=Φ 'FTΦFTEigenvalue Arithmetic root, verifies all eigenvalue λ i ∈ (1- δ of G (Φ F, T)K,1+δK), i=1 ..., T, then ΦFMeet RIP, and by asking SolutionOptimization problem reconstructs original signal, that is, BP reconstructing original signal by linear programming method Algorithm；For actual compression signal, the such as collection of voice signal, then Φ is changedFFor following form：
If signal conversion basic matrix Ψ on have it is openness, By solving such as following formula optimization problem, Accurate Reconstruction goes out original signal：Wherein Φ Uncorrelated to Ψ, Ξ is referred to as CS matrixes；
With voice acquisition module by input module wired connection, for being converted into the signal conversion of the signal of telecommunication to the voice for gathering Module；
With signal conversion module wired connection, for being analyzed the single-chip microcomputer of process to the signal of telecommunication；The single-chip microcomputer is used to obtain Take the signal of signal conversion module transmission and process is amplified to signal；Average, variance, signal are extracted from every segment signal Accumulated value and the basic time domain parameter of peak value 4, determine whether doubtful by the difference of 4 parameter values of adjacent segment signal The ground floor decision-making that the non-type situation of sound occurs judges：The no person of wavelet packet denoising is down performed if having, is jumped to and is performed acquisition The signal of signal conversion module transmission；Wavelet packet denoising is carried out to the signal for gathering using Wavelet Packet Algorithm is improved；Using improvement Wavelet Packet Algorithm carries out WAVELET PACKET DECOMPOSITION and reconstruct to the signal for gathering, and obtains list band reconstruction signal；From the list band of reconstruct Extract in signal：Time domain energy, time domain peak, frequency domain energy, frequency domain peak value, coefficient of kurtosis, variance, frequency spectrum and coefficient of skewness 8 The individual parameter for representing signal characteristic；Using principal component analytical method, select 3 to 8 substantially expression sound can send out from above-mentioned parameter Penetrating the parameter composition characteristic vector of signal characteristic, and these characteristic vectors are input to into support vector machine carries out decision-making judgement, i.e., Second layer decision-making judgement, determines whether that non-type situation occurs according to the output of support vector machine；
With single-chip microcomputer wired connection, for storing the RAM memory and mram memory of Japanese RP；
With single-chip microcomputer wired connection, the enquiry module inquired about for the Japanese of user to be pronounced；
With single-chip microcomputer wired connection, for the pronunciation of the Japanese of user and the pronunciation of the Japanese of standard to be analyzed the analysis mould of contrast Block；The analysis module is used for construction multilevel iudge matrix two-by-two, according to 1~9 scaling law successively to each key element between any two Carry out importance degree assignment, Judgement Matricies U=(uij)n×n, wherein uijExpression factor uiAnd ujRelative to RP letter The importance value of breath, matrix U has property：uii=1, uij=1/uji, i, j=1,2 ..., n draw judgment matrix；By matrix X1 ~X5By row normalization, i.e.,：Calculating matrix Y is： Y matrix by rowss are added, by formulaDraw：W1=(2.652 0.686 0.253 0.409)T, W2=(1 1)T, W3=(1.273 0.371 0.221 2.135)T, W4=(1.9 0.319 0.781)T, W5=(2.121 0.604 0.275)T；Obtain and vector is normalized, by formulaObtain weight vector： Calculate the Maximum characteristic root λ of judgment matrixmax, by formula Calculate: According to FormulaConsistency check is carried out, is obtained：CI1=0.019, CI2=0, CI3= 0.031, CI4=0.020, CI5=0.048；By formulaObtain CR1=0.022, CR2=0, CR3=0.035, CR4= 0.038, CR5=0.092；CR ＜ 0.1, are satisfied by coherence request, therefore the relative weighting of each factor Relative analyses go out the Japanese pronunciation of user and the Japanese of standard pronounces after consistency check Contrast；
With single-chip microcomputer wired connection, for the processing module that the character of the nonstandard pronunciation of user is marked and is processed；It is described Processing module is used to receive safeguards more new command；The nonstandard pronunciation information of user and need are obtained according to the maintenance more new command Safeguard the dimension table information of renewal；According to the dimension table configuration information that the dimension table acquisition of information pre-sets；Wherein, institute State in dimension table configuration information and safeguard that the source database at the dimension table place for updating, the dimension table need together with the needs The purpose data base of step and dimension table operating right information；According to the nonstandard pronunciation information of the user and the dimension table Operating right information, judges whether the nonstandard pronunciation information of the user meets the dimension table operating right information；If described The nonstandard pronunciation information of user meets the dimension table operating right information, needs to safeguard that the dimension table for updating is carried out more to described New operation；The dimension table being updated after operation is synchronized in the RAM memory and mram memory；
With single-chip microcomputer by data update module wired connection, for being updated the external memory of Japanese pronunciation；
With single-chip microcomputer wired connection, for carrying out the network connecting module of wireless network connection；
Wirelessly it is connected by GPRS wireless networks with network connecting module, what the Japanese for user error to be pronounced was stored Cloud Server.
2. Japanese pronunciation correction system as claimed in claim 1, it is characterised in that wavelet packet denoising and WAVELET PACKET DECOMPOSITION and weight Structure includes：
Signals extension, to each layer signal of WAVELET PACKET DECOMPOSITION horizontal parabola continuation is entered；
If signal data is x (a), x (a+1), x (a+2), then the expression formula of continuation operator E is：
x ( a - 1 ) = 3 x ( a ) - 3 x ( a + 1 ) + x ( a + 2 ) x ( a + 3 ) = 3 x ( a + 2 ) - 3 x ( a + 1 ) + x ( a ) - - - ( 1 ) ;
Eliminate list band un-necessary frequency composition；
By the signal after continuation and decomposition low pass filter h0Convolution, obtains low frequency coefficient, then at HF-cut-IF operators Reason, removes unnecessary frequency content, then carries out down-sampling, obtains next layer of low frequency coefficient；By the signal after continuation and decomposition High pass filter g0Convolution, obtains high frequency coefficient, then through the process of LF-cut-IF operators, removes unnecessary frequency content, then Down-sampling is carried out, next layer of high frequency coefficient is obtained, shown in HF-cut-IF operators such as formula (2), LF-cut-IF operators such as formula (3) institute Show；
X ( k ) = Σ n = 0 N j - 1 x ( n ) W k n , 0 ≤ k ≤ N j 4 ; 3 N j 4 ≤ k ≤ N j X ( k ) = 0 , x ( n ) = Σ k = 0 N j - 1 x ( k ) W - k n , - - - ( 2 )
X ( k ) = Σ n = 0 N j - 1 x ( n ) W k n , N j 4 ≤ k ≤ 3 N j 4 X ( k ) = 0 , x ( n ) = Σ k = 0 N j - 1 x ( k ) W - k n , - - - ( 3 )
In (2), (3) formula, x (n) is 2jThe coefficient of wavelet packet, N on yardstickjRepresent 2jThe length of data on yardstick,K=0,1 ..., Nj-1；N=0,1 ..., Nj-1；
It is described to obtain list band reconstruction signal, specifically include：The high and low frequency coefficient for obtaining is up-sampled, then respectively with High pass reconstruction filter g1With low-pass reconstruction filter h1Convolution, the signal for obtaining is calculated respectively with HF-cut-IF, LF-cut-IF Subprocessing, obtains list band reconstruction signal.
3. Japanese pronunciation correction system as claimed in claim 1, it is characterised in that in the processing module, dimension table handling Authority information includes：The nonstandard pronunciation information mark of user with dimension table operating right；
It is described to judge whether the nonstandard pronunciation information of the user meets the dimension table operating right information, including：
Judge the nonstandard pronunciation information of the user whether in the nonstandard pronunciation letter of the user with dimension table operating right In breath mark；
The maintenance more new command is for increase content instruction, change content instruction or deletes content instruction；
Before operation is updated to the dimension table for needing maintenance to update, including：
According to the maintenance more new command, it is determined that needs safeguard the field for updating, and the word for needing to safeguard renewal is got Segment identification；
The field configuration information for pre-setting is got according to the field identification and the dimension table configuration information；Wherein, The field configuration information includes field contents ordering rule, field ordering information, the field restrictive condition of the field；
If the maintenance more new command is to increase content instruction, described to need to safeguard that the dimension table for updating is updated behaviour to described Make, including：
Obtain the corresponding batch data content of the increase content instruction；
According to the batch data content, in one or more fields in the dimension table field contents are increased；
According to the field contents ordering rule, the field contents are ranked up；
According to the field ordering information, each field in dimension table is ranked up；
If the maintenance more new command is change content instruction, described to need to safeguard that the dimension table for updating is updated behaviour to described Make, including：
Obtain the corresponding batch data content of the change content instruction；
According to the batch data content, in one or more fields in the dimension table field contents are changed；
If the maintenance more new command is to delete content instruction, described to need to safeguard that the dimension table for updating is updated behaviour to described Make, including：
Field contents are deleted in one or more fields in the dimension table；
Also include：
Judge whether each field after the increase field contents, change field contents or deletion field contents meets described Field restrictive condition；
If there is field to be unsatisfactory for the field restrictive condition, information is generated；The information is unsatisfactory for institute for prompting The Field Count of field restrictive condition is stated, and prompting is unsatisfactory for the field relevant information of the field restrictive condition；The field phase Pass information includes the field identification or field name of the field.
4. Japanese pronunciation correction system as claimed in claim 1, it is characterised in that the single-chip microcomputer and the wired company of power module Connect, for providing power supply.
5. Japanese pronunciation correction system as claimed in claim 1, it is characterised in that the single-chip microcomputer is by output module and raises Sound device wired connection, for the output of standard Japanese pronunciation.
6. Japanese pronunciation correction system as claimed in claim 1, it is characterised in that the single-chip microcomputer is by output module and aobvious Display screen wired connection, is shown for the Japanese after labelling and process.
CN201611218260.1A 2016-12-26 2016-12-26 System for Japanese pronunciation correction Pending CN106682161A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611218260.1A CN106682161A (en) 2016-12-26 2016-12-26 System for Japanese pronunciation correction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611218260.1A CN106682161A (en) 2016-12-26 2016-12-26 System for Japanese pronunciation correction

Publications (1)

Publication Number Publication Date
CN106682161A true CN106682161A (en) 2017-05-17

Family

ID=58870630

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611218260.1A Pending CN106682161A (en) 2016-12-26 2016-12-26 System for Japanese pronunciation correction

Country Status (1)

Country Link
CN (1) CN106682161A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107481562A (en) * 2017-09-11 2017-12-15 芜湖凡达机械科技有限公司 A kind of Japanese pronunciation correction device
CN108766059A (en) * 2018-05-21 2018-11-06 重庆交通大学 A kind of cloud service English teaching equipment and teaching method

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103036573A (en) * 2012-12-03 2013-04-10 南京工业大学 Compressed sensing signal collection method based on filtering
CN103310666A (en) * 2013-05-24 2013-09-18 深圳市九洲电器有限公司 Language learning device
CN103366123A (en) * 2013-05-07 2013-10-23 天津大学 Software risk assessment method based on defect analysis
CN104252452A (en) * 2013-06-25 2014-12-31 腾讯科技（深圳）有限公司 Data management method and device
CN106556760A (en) * 2016-11-16 2017-04-05 北华大学 A kind of automated test control system of the vehicle mounted multimedia based on cloud computing

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103036573A (en) * 2012-12-03 2013-04-10 南京工业大学 Compressed sensing signal collection method based on filtering
CN103366123A (en) * 2013-05-07 2013-10-23 天津大学 Software risk assessment method based on defect analysis
CN103310666A (en) * 2013-05-24 2013-09-18 深圳市九洲电器有限公司 Language learning device
CN104252452A (en) * 2013-06-25 2014-12-31 腾讯科技（深圳）有限公司 Data management method and device
CN106556760A (en) * 2016-11-16 2017-04-05 北华大学 A kind of automated test control system of the vehicle mounted multimedia based on cloud computing

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107481562A (en) * 2017-09-11 2017-12-15 芜湖凡达机械科技有限公司 A kind of Japanese pronunciation correction device
CN108766059A (en) * 2018-05-21 2018-11-06 重庆交通大学 A kind of cloud service English teaching equipment and teaching method

Similar Documents

Publication Publication Date Title
CN106611169B (en) A kind of dangerous driving behavior real-time detection method based on deep learning
Schomaker Crown-condition classification: a guide to data collection and analysis
CN106796716B (en) For providing the device and method of super-resolution for low-resolution image
CN105808970A (en) Online cognitive assessment system and assessment method
CN108874959A (en) A kind of user's dynamic interest model method for building up based on big data technology
CN105654414A (en) Urban multi-disaster risk loss evaluation system based on open source system framework and building spatial database and method thereof
DE112016004922T5 (en) Construction and analysis system of the touch screen user button behavior mode and its identity recognition method
CN106202377A (en) A kind of online collaborative sort method based on stochastic gradient descent
Zhang et al. Bayesian analysis of climate change effects on observed and projected airborne levels of birch pollen
CN105931153A (en) Indirect questionnaire assessment method based on neural network prediction analysis model
CN106682161A (en) System for Japanese pronunciation correction
CN113408957B (en) Classroom teaching evaluation method based on combined weighting method
CN107784379A (en) A kind of application for studying abroad forecasting system and method
CN107301409A (en) Learn the system and method for processing electrocardiogram based on Wrapper feature selectings Bagging
CN112836937A (en) Flood disaster loss evaluation method based on entropy weight and BP neural network technology
CN114566282B (en) Treatment decision system based on echocardiogram detection report
Hallett et al. Citizen science and tree health assessment: how useful are the data
CN111553593A (en) Safety risk assessment method and system for urban street tree
CN109740418B (en) Yoga action identification method based on multiple acceleration sensors
CN111210912A (en) Parkinson prediction method and device
CN105260963A (en) Subject accomplishment evaluation system
CN110363833A (en) A kind of complete human body sport parameter representation method based on local rarefaction representation
CN112712114A (en) Instrument analysis method, system, equipment and medium based on TextCNN-BilSTM
CN107845042A (en) Supervisory systems is imposed in house in a kind of state-owned land
DE112020007405T5 (en) Method for finding a community based on an uncertain graph

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20170517

RJ01 Rejection of invention patent application after publication