CN110164473A - A kind of chord arrangement detection method based on deep learning - Google Patents
A kind of chord arrangement detection method based on deep learning Download PDFInfo
- Publication number
- CN110164473A CN110164473A CN201910422361.8A CN201910422361A CN110164473A CN 110164473 A CN110164473 A CN 110164473A CN 201910422361 A CN201910422361 A CN 201910422361A CN 110164473 A CN110164473 A CN 110164473A
- Authority
- CN
- China
- Prior art keywords
- chord
- deep learning
- detection method
- note
- detection
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 46
- 238000013135 deep learning Methods 0.000 title claims abstract description 21
- 238000000605 extraction Methods 0.000 claims abstract description 5
- 238000000513 principal component analysis Methods 0.000 claims abstract description 5
- 238000000034 method Methods 0.000 claims description 23
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 abstract description 3
- 230000002708 enhancing effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 230000000116 mitigating effect Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 241001255907 Tonica Species 0.000 description 1
- 210000003484 anatomy Anatomy 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- 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
- G09B15/00—Teaching music
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Signal Processing (AREA)
- Acoustics & Sound (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Educational Technology (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Auxiliary Devices For Music (AREA)
Abstract
A kind of chord based on deep learning disclosed by the invention arranges detection method, is related to chord detection technique field.The chord arranges detection method, pass through the algorithm of application deep learning, feature extraction is carried out to part polyphonic ring tone, its dimensional information is further compressed using Principal Component Analysis, whether there is mistake to chord by SVM classifier and type of error judges and classification, and the note in density array mistake string sound is positioned using algorithm of target detection, pass through the prediction frame of adjacent part later, it measures its pixel distance and is scaled interval degree, the finally arrangement rule based on chord, decision error information, and the note for mistake occur is labeled.A kind of chord arrangement detection method based on deep learning disclosed by the invention alleviates teacher to the workload to check up on students' homework by way of student resource inspection, while having also achieved the purpose of enhancing learning efficiency.
Description
Technical field
The present invention relates to chord detection technique fields, and in particular to a kind of chord arrangement detection side based on deep learning
Method.
Background technique
The course of music speciality is divided into practice course and theoretical course at present, " music theory " in music speciality theory course,
" and acoustics ", " polyphony " are directed to the writing of " harmony (chord) ".
Domestic and acoustics study course generallys use " four harmony " --- in the joint being made of trble staff and bass staff
There are four part, two parts of trble staff (high line and alto), two part (tenors of bass staff altogether in stave
And basso).Three or more notes are combined in the longitudinal direction according to three degree of stacked relation, just become chord,
This is the vertical structure of general four harmony.In the writing of four harmony, common chords have it is intensive with two kinds of open ranking methods,
Seventh chord has intensive, open and mixing three kinds of ranking methods.
In recent years, with the continuous development of computer technology, China's major part institution of higher learning all actively push forward digitlization,
Interactive teaching pattern.But domestic music teaching harmony authoring system it is not perfect, it is existing be used to write chord and
Sound, polyphony are all some staffs introduced from foreign countries are set the chessman on the chessboard according to the chess manual software, as Sibelius (Xi Beiliusi), Tonica,
Overture and Finale etc., these softwares are only capable of writing music score, do not have the analysis to harmony, detection, determine and illustrate function
Energy.Chord arrangement detection is one of the Center Technology of music teaching harmony authoring system detection unit, it ties the longitudinal direction of harmony
The writing result judgement of structure --- " chord " plays an important role.It mainly include two sides in the detection arranged chord
The research contents in face: first is that the direction determining note affiliated part dry according to symbol, detects the presence of and part intersection occurs;Second is that detection
The ranking method of chord.These Detection tasks are currently manually to be completed by teacher, the intelligent measurement that is arranged using computer chord
The technology country is still blank.
Therefore, in view of problem above, it is necessary to propose a kind of new chord arrangement detection method, pass through student resource inspection
Mode can also reinforce the efficiency of students'autonomous study while mitigating teachers ' teaching pressure.
Summary of the invention
In view of this, the present invention discloses a kind of chord arrangement detection method based on deep learning, it is powerful using computer
Calculation function, the part that is likely to occur in chord is intersected by deep learning algorithm and chord arrangement problems judge
And classification, and carry out the mark of error message for the chord for density array problem occur and revise one's view, certainly by student
The mode of main inspection alleviates teacher to the workload to check up on students' homework, while having also achieved the purpose of enhancing learning efficiency.
A kind of chord based on deep learning that purpose according to the present invention proposes arranges detection method, including following step
It is rapid:
Step 1: feature extraction is carried out to polyphonic ring tone using convolutional neural networks, using Principal Component Analysis to neuron
The characteristic dimension information that network extracts further is compressed, and is classified by SVM to compressed polyphonic ring tone feature, with this
Whether there is mistake to chord and type of error is classified and determined.
Step 2: three parts above the chord of density array mistake are carried out with the positioning of prediction frame, to the position of note
It sets and is accurately positioned.
Step 3: the interval degree above measurement chord between three parts.
Step 4: the interval degree of adjacent part in ranking method and the part of top three based on chord, to having write and
The ranking method of string carries out detection judgement.
Step 5: to detecting that the wrong chord of ranking method is labeled, and explanatory note is provided.
Preferably, in step 2, music score is divided into the subgraph of S × S, chord detection, and antithetical phrase are carried out to each subgraph
The position of chord note is accurately positioned in figure.
Preferably, in step 3, using chord part detection algorithm, the part of top three in chord is positioned,
The coordinate that frame is predicted by part, measures the pixel distance between adjacent part.
Preferably, in step 4, be scaled interval degree using part pixel distance, based on chord ranking method rule to
String carries out detection judgement.
Preferably, in step 5, the notation methods of wrong chord are the red note for being different from normally black note.
Compared with prior art, the advantages of a kind of chord based on deep learning disclosed by the invention arranges detection method
It is:
(1) this method based on the technology of deep learning to chord arrange carry out intellectual analysis, detection, to correct result into
Row summary description is labeled the mistake of harmony writing, while mitigating teachers ' teaching pressure, can also reinforce student certainly
Primary learning efficiency.
(2) this method, which not only solves, artificially judges error brought by music score error message subjectivity, is also equipped with raising
Determine occur the advantages of wrong chord positional accuracy and rapidity in music score.
Detailed description of the invention
It, below will be to embodiment or existing for the clearer technical solution for illustrating the embodiment of the present invention or the prior art
Attached drawing needed in technical description does simple introduction, it is clear that, the accompanying drawings in the following description is only this hair
Bright some embodiments, it will be clear to the skilled person that without creative efforts, may be used also
Other accompanying drawings can also be obtained according to these attached drawings.
Fig. 1 is that the present invention carries out error message mark schematic diagram to the chord for density array mistake occur;
Fig. 2 is the flow chart that the chord disclosed by the invention based on a kind of deep learning arranges detection method.
Specific embodiment
A specific embodiment of the invention is described briefly with reference to the accompanying drawing.Obviously, described embodiment is only
It is a part of the embodiments of the present invention, rather than whole embodiments, based on the embodiments of the present invention, ordinary skill people
Member's every other embodiment obtained without making creative work, belongs to the scope of protection of the invention.
Fig. 2 shows preferred embodiments of the present invention by Fig. 1-, have carried out detailed anatomy to it from different angles respectively.
A kind of chord based on deep learning as shown in Figs. 1-2 arranges detection method, passes through convolutional neural networks first,
The feature of each part polyphonic ring tone, such as the direction that the shape of symbol head, symbol are dry are tentatively extracted, using Principal Component Analysis Algorithm to feature
Dimensional information is further compressed, and is then classified using SVM to compressive features map, is determined whether chord density occurs with this
Arrangement or part intersect mistake;Using algorithm of target detection, it is accurately fixed to carry out to the polyphonic ring tone position for density array mistake occur
Position;Then to its top, three parts carry out the positioning of prediction frame, and by the coordinate of prediction frame measure adjacent part it
Between pixel distance;Interval degree is scaled by the pixel distance of adjacent part;Finally based on chord ranking method rule to
String carries out detection judgement, carries out related mark explanation to error message.It is specific as follows:
Step 1: feature extraction is carried out to polyphonic ring tone using convolutional neural networks, using Principal Component Analysis to neuron
The characteristic dimension information that network extracts further is compressed, and is classified finally by SVM to compressed polyphonic ring tone feature,
Whether there is mistake to chord with this and type of error is classified and determined.
Step 2: being based on deep learning method, the chord for density array mistake occur is accurately positioned.By that will find pleasure in
Spectrogram picture is divided into the subgraph of S × S, and the neuron in each subgraph is responsible for carrying out target detection to the object fallen into the cell
And positioning is arranged each subgraph and contains up to the sliding window of three prediction objects due to the particularity of music score, and is indicated with B.
The information of each sliding window is by coordinate (Tx,Ty,Tw,Th) and confidence level C composition, the frame that wherein neuroid detects
Bias size is by (Tx,Ty) indicate, detect the width of frame, high and input picture ratio column use T respectivelywAnd ThIt indicates, confidence level
C indicates a possibility that there are chords in detection frame size, its calculation formula is: C=PO+PIOU, wherein POIndicate sliding window
In include chord object probability, PIOUThe overlapping area of sliding window and test object region is indicated, if the sliding window of detection
In contain chord then PO=1, it is otherwise PO=0.Due to only detecting the chord in music score, then classification C=1 is detected, then final nerve
The output dimension of metanetwork are as follows: S × S × (B × 5+C).
Loss function is defined to be missed by detection frame error of coordinate, the sliding window confidence level error of each subgraph and classification
Difference determines.Wherein detect error of coordinate function, confidence level error function, error in classification function and the overall error function of frame
Definition is as follows respectively:
losstotal=losscoord+lossconf+lossclass;
Wherein IijIndicate whether contain target, S in j-th of sliding window in i-th of subgraph2It indicates in a music score figure
The number of detection subgraph is carried out, B indicates the number of sliding window in each subgraph.λcoordFor the weight system of error of coordinate
Number, xi,yiRespectively indicate the central point abscissa and ordinate of sliding window in i-th of subgraph, wiAnd hiFor the width for predicting frame
Degree and length,WithRespectively indicate the central point transverse and longitudinal coordinate and width of the sliding window of true i-th of subgraph
And length.For the balance being better balanced between coordinate prediction error and objective degrees of confidence score, λ is introducednoobjAs power
Weight coefficient so that in the picture without target frame when, reduce to confidence level punishment dynamics.CiRepresent setting in true detection block
Reliability,The confidence level containing target is predicted whether in i-th of subgraph.It indicates to predict the C classification in i-th of subgraph
Probability.
Step 3: interception has detected that the part of polyphonic ring tone, it is accurately fixed to the progress again of its three part position in top
Position.The prediction frame for defining part is (Lx,Ly,Lw,Lh), the center point coordinate of frame is (Lx,Ly), the width of frame and high difference
For Lw、Lh, then the center point coordinate of its upper side frame is Lup=(Lx,Ly-Lh/ 2), the center point coordinate of lower frame is Ldown=(Lx,
Ly+Lh/ 2), then any two symbol head can be with respect to the distance between part is defined as:
It pel spacing and is changed step 4: the ranking method rule based on chord, between the adjacent part in calculated top three
Calculating is interval degree, and then carries out detection judgement to the ranking method for having write chord, judges three parts in chord top in music score
Arrangement it is whether correct.
It is marked step 5: being given to the chord not being inconsistent normally, and reason and how modifying of its mistake is referred to
Show explanation.Compared with normal black note, the note for writing mistake can be noted as red.Operator can tie according to judgement
Fruit voluntarily thinks deeply amendment scheme, and also the property of can choose provides instruction, above chord provides explanatory note.As shown in Figure 1, one
Common chords detect that the ranking method of chord is wrong, and entire chord (four sounds) is all noted as red, click this chord, on chord
Side provides explanatory note.
By above five steps, can to chord in music score, whether mistake judges, and there is mistake to chord
Music score carries out information labeling.
In conclusion a kind of chord based on deep learning disclosed by the invention arranges detection method, it is based on deep learning
Technology chord arranged carry out intellectual analysis, detection, summarize explanation to correct result, to harmony writing it is wrong into
Rower note, while mitigating teachers ' teaching pressure, can also reinforce students'autonomous study efficiency.Meanwhile this method not only solves
It has determined error brought by artificial judgement music score error message subjectivity, has been also equipped with raising and determines occur wrong chord position in music score
The advantages of setting accuracy and rapidity.
The foregoing description of the disclosed embodiments makes professional and technical personnel in the field can be realized and use the present invention.
Various modifications to these embodiments mode will be readily apparent to those skilled in the art, and determine herein
The General Principle of justice can be realized in other embodiments without departing from the spirit and scope of the present invention.Therefore, originally
Invention is not intended to be limited to the embodiments shown herein, and is to fit to the principles and novel features disclosed herein phase
Consistent widest scope.
Claims (5)
1. a kind of chord based on deep learning arranges detection method, which comprises the following steps:
Step 1: feature extraction is carried out to polyphonic ring tone using convolutional neural networks, using Principal Component Analysis to neuroid
The characteristic dimension information of extraction is further compressed, and is classified by SVM to compressed polyphonic ring tone feature, with this to
Whether string there is mistake and type of error is classified and determined;
Step 2: carrying out the positioning of prediction frame to three parts above the chord of density array mistake, to the position of note into
Row is accurately positioned;
Step 3: the interval degree above measurement chord between three parts;
Step 4: the interval degree of adjacent part in ranking method and the part of top three based on chord, to having write chord
Ranking method carries out detection judgement;
Step 5: to detecting that the wrong chord of ranking method is labeled, and explanatory note is provided.
2. a kind of chord based on deep learning according to claim 1 arranges detection method, which is characterized in that step 2
In, music score is divided into the subgraph of S × S, chord detection is carried out to each subgraph, and carry out in the subgraph and position of string note
It is accurately positioned.
3. a kind of chord based on deep learning according to claim 1 arranges detection method, which is characterized in that step 3
In, using chord part detection algorithm, the part of top three in chord is positioned, the seat of frame is predicted by part
Mark, measures the pixel distance between adjacent part.
4. a kind of chord based on deep learning according to claim 3 arranges detection method, which is characterized in that step 4
In, it is scaled interval degree using part pixel distance, detection judgement is carried out to chord based on chord ranking method rule.
5. a kind of chord based on deep learning according to claim 1 arranges detection method, which is characterized in that step 5
In, the notation methods of wrong chord are the red note for being different from normally black note.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910422361.8A CN110164473B (en) | 2019-05-21 | 2019-05-21 | Chord arrangement detection method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910422361.8A CN110164473B (en) | 2019-05-21 | 2019-05-21 | Chord arrangement detection method based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110164473A true CN110164473A (en) | 2019-08-23 |
CN110164473B CN110164473B (en) | 2021-03-26 |
Family
ID=67631637
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910422361.8A Expired - Fee Related CN110164473B (en) | 2019-05-21 | 2019-05-21 | Chord arrangement detection method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110164473B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111968452A (en) * | 2020-08-21 | 2020-11-20 | 江苏师范大学 | Harmony learning method and device and electronic equipment |
CN112381792A (en) * | 2020-11-13 | 2021-02-19 | 中国人民解放军空军工程大学 | Radar wave-absorbing coating/electromagnetic shielding film damage intelligent imaging online detection method based on deep learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008040283A (en) * | 2006-08-09 | 2008-02-21 | Kawai Musical Instr Mfg Co Ltd | Chord name detecting device and program for chord name detection |
CN102723079A (en) * | 2012-06-07 | 2012-10-10 | 天津大学 | Music and chord automatic identification method based on sparse representation |
CN103714806A (en) * | 2014-01-07 | 2014-04-09 | 天津大学 | Chord recognition method combining SVM with enhanced PCP |
CN106446952A (en) * | 2016-09-28 | 2017-02-22 | 北京邮电大学 | Method and apparatus for recognizing score image |
-
2019
- 2019-05-21 CN CN201910422361.8A patent/CN110164473B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008040283A (en) * | 2006-08-09 | 2008-02-21 | Kawai Musical Instr Mfg Co Ltd | Chord name detecting device and program for chord name detection |
CN102723079A (en) * | 2012-06-07 | 2012-10-10 | 天津大学 | Music and chord automatic identification method based on sparse representation |
CN103714806A (en) * | 2014-01-07 | 2014-04-09 | 天津大学 | Chord recognition method combining SVM with enhanced PCP |
CN106446952A (en) * | 2016-09-28 | 2017-02-22 | 北京邮电大学 | Method and apparatus for recognizing score image |
Non-Patent Citations (2)
Title |
---|
刘婷: "《基于相空间重构和支持向量机的和弦识别》", 《计算机与数字工程》 * |
王蒙蒙等: "《基于鲁棒音阶特征的测度学习SVM的音乐和弦识别》", 《信号处理》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111968452A (en) * | 2020-08-21 | 2020-11-20 | 江苏师范大学 | Harmony learning method and device and electronic equipment |
CN112381792A (en) * | 2020-11-13 | 2021-02-19 | 中国人民解放军空军工程大学 | Radar wave-absorbing coating/electromagnetic shielding film damage intelligent imaging online detection method based on deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN110164473B (en) | 2021-03-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109919934A (en) | A kind of liquid crystal display panel defect inspection method based on the study of multi-source domain depth migration | |
CN109543742A (en) | A kind of image local information transfer method based on GAN and Self-Attention | |
CN110414380A (en) | A kind of students ' behavior detection method based on target detection | |
CN110164473A (en) | A kind of chord arrangement detection method based on deep learning | |
CN115409069A (en) | Village and town building identification method, classification method, device, electronic equipment and medium | |
CN115239527A (en) | Teaching behavior analysis system for teaching characteristic fusion and modeling based on knowledge base | |
Ellison | Intelligent accountability: Re-thinking the concept of “accountability” in the popular discourse of education policy | |
CN109272160A (en) | Score on Prediction system and prediction technique | |
CN102646164A (en) | Land use change modeling method and system implemented in combination with spatial filtering | |
CN109948609A (en) | Intelligently reading localization method based on deep learning | |
CN116757347B (en) | Railway line selection method and system based on deep learning | |
CN107967540A (en) | Student's academic warning system and method | |
CN109145956A (en) | Methods of marking, device, computer equipment and storage medium | |
CN111611854B (en) | Classroom condition evaluation method based on pattern recognition | |
Thomas | An analysis of the research methodology of the Ramirez study | |
CN115620143A (en) | New classical architecture style identification system, construction method and identification method | |
Chew et al. | Interactive multi-scale visualizations of tonal evolution in MuSA. RT Opus 2 | |
Hossain et al. | History and milestones of higher education in India | |
CN112668771A (en) | Comprehensive evaluation site selection method and system for submarine tunnel | |
CN110309727A (en) | A kind of foundation of Building recognition model, Building recognition method and apparatus | |
Muraina et al. | Forecasting Students’ Job Placement using Data Science Paradigms | |
CN112597948B (en) | Urban land utilization change prediction method | |
Zhao | Study on the effectiveness of the asr-based english teaching software in helping college students’ listening learning | |
Krishna et al. | An Efficient Method to Predict Student Education Performance using GoogleNet compared with InceptionNet | |
CN108764537A (en) | A kind of multi-source community label prediction of the development trend method based on A-TrAdaboost algorithms |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210326 |
|
CF01 | Termination of patent right due to non-payment of annual fee |