CN107256710A - A kind of humming melody recognition methods based on dynamic time warp algorithm - Google Patents
A kind of humming melody recognition methods based on dynamic time warp algorithm Download PDFInfo
- Publication number
- CN107256710A CN107256710A CN201710648569.2A CN201710648569A CN107256710A CN 107256710 A CN107256710 A CN 107256710A CN 201710648569 A CN201710648569 A CN 201710648569A CN 107256710 A CN107256710 A CN 107256710A
- Authority
- CN
- China
- Prior art keywords
- humming
- melody
- song
- dynamic time
- time warp
- 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
Links
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; 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
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; 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
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; 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
- G10L25/78—Detection of presence or absence of voice signals
- G10L25/87—Detection of discrete points within a voice signal
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2210/00—Aspects or methods of musical processing having intrinsic musical character, i.e. involving musical theory or musical parameters or relying on musical knowledge, as applied in electrophonic musical tools or instruments
- G10H2210/031—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal
- G10H2210/041—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal based on mfcc [mel -frequency spectral coefficients]
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2210/00—Aspects or methods of musical processing having intrinsic musical character, i.e. involving musical theory or musical parameters or relying on musical knowledge, as applied in electrophonic musical tools or instruments
- G10H2210/031—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal
- G10H2210/066—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal for pitch analysis as part of wider processing for musical purposes, e.g. transcription, musical performance evaluation; Pitch recognition, e.g. in polyphonic sounds; Estimation or use of missing fundamental
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2210/00—Aspects or methods of musical processing having intrinsic musical character, i.e. involving musical theory or musical parameters or relying on musical knowledge, as applied in electrophonic musical tools or instruments
- G10H2210/031—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal
- G10H2210/071—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal for rhythm pattern analysis or rhythm style recognition
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Auxiliary Devices For Music (AREA)
Abstract
The invention discloses a kind of humming melody recognition methods based on dynamic time warp algorithm, realize and the humming melody of humming person is fast and accurately recognized.Humming melody audio signal is pre-processed first, noise reduction process is carried out to voice signal, the mixed and disorderly sound in signal is removed.Then preemphasis and adding window sub-frame processing are carried out to the signal after noise reduction, obtains high-quality voice signal.The operation such as end-point detection and the extraction of MFCC coefficients is carried out to voice signal, the parameter extraction for representing voice signal substantive characteristics is come out to the characteristic sequence for obtaining humming input melody.Characteristic sequence will be obtained and the characteristic sequence of the song in library is contrasted, a part of song is deleted according to melody characteristicses and editing distance.Using improved dynamic time warp algorithm, the characteristic sequence of remaining song is accurately matched with target signature sequence, candidate song is obtained, realize the identification of humming melody.
Description
Technical field
The present invention relates to a kind of humming melody recognition methods based on dynamic time warp algorithm, belong to speech recognition neck
Domain.
Background technology
In content-based music information retrieval research, humming melody recognition methods is its core algorithm and research weight
Point, the performance of recognition methods has direct influence to recognition efficiency and recognition result.It is unfamiliar with to song title or singer
Situation, by hum melody recognize we can with it is very convenient efficiently according to individual humming inquire respective songs.Simultaneously
It also has quite varied application in terms of the tune analysis of song and the performance level marking of singer.Know in humming melody
In other method, we should consider to hum note increase and decrease and pitch bias phenomenon in voice, consider that user can be in song again
Bent arbitrary period is hummed.Therefore use appropriate recognition strategy and matching process particularly significant.
The content of the invention
It is an object of the invention to provide a kind of humming melody recognition methods based on dynamic time warp algorithm, it can have
The voice messaging that effect ground is hummed according to humming person identifies the information of song.The present invention uses following technical scheme:
A kind of humming melody recognition methods based on dynamic time warp algorithm, comprises the following steps:
(1) voice messaging is pre-processed.The voice messaging of humming is subjected to denoising, the operation such as preemphasis and adding window framing is carried
The quality of high voice signal, making the HFS of signal becomes flat, and voice signal is integrally steady, is easy to follow-up analysis;
(2) feature extraction.End-point detection is carried out to the voice signal treated by above-mentioned steps and MFCC coefficients are extracted
Deng operation, the parameter extraction for representing voice signal substantive characteristics can be come out, obtain the characteristic sequence of humming input melody;
(3) the quick screening based on melody characteristicses.This method is used as feature melody according to half beat of song.Basis first
The pitch of song, which rises and falls, calculates the descant and contrabass half beat of a song.With melody to be measured half in delete target music libraries
Beat differs larger song.By this step, a part of song is quickly deleted, melody identification process is improved;
(4) the quick screening based on editing distance.The characteristic sequence for inputting melody will be hummed first poor according to pitch, conversion
Pass through into character string sequence, with target music libraries and remaining song progress calculating editor is screened based on melody characteristicses Rapid matching
Distance.Reject the song that distance differs larger;
(5) the accurate identification based on improved dynamic time warp algorithm.In humming identification, it will simply can not groan
Sing input melody characteristic sequence be compared with target template sequence because humming person there may be in humming it is various
Humming mistake, such as addition note deletes note etc., therefore it is particularly significant to carry out time planning to template.Dynamic time warp
Algorithm is by constantly calculating the distance of two vectors to find optimal coupling path, so two obtained vectors are cumulative distances
Minimum regular integer, it ensure that there are maximum acoustics similar features between them.But dynamic time warp algorithm is deposited
It is slow in retrieval time, the problems such as required memory space is big, therefore this method relaxes from end points regarding to the issue above, across sentence retrieval,
Several aspects such as cost function are improved to dynamic time warp algorithm.Realized by the dynamic time warp algorithm after improvement
The accurate identification of humming melody.
Advantages of the present invention:
1st, invention introduces across sentence retrieval and tail end point relaxation, the time required to reducing dynamic time warp algorithm
20%.
2nd, the present invention introduces the feature of the duration of a sound in the cost function calculation of dynamic time warp algorithm, hit rate is existed
5% is improved on the basis of original.
3rd, the present invention do not limit user humming mode or rhythm it is less accurate in the case of, have compared to former algorithm
More preferable robustness.
4th, the present invention is retrieved for traditional dynamic time warp algorithm across sentence, is changed in terms of end points relaxation
Enter, and otonaga features are introduced in the cost function calculation of algorithm, the property of humming melody recognizer is improved on the whole
Can, improve recognition efficiency.
Brief description of the drawings
Fig. 1 is flow chart of the method for the present invention.
Fig. 2 is dynamic time warp algorithm principle figure of the present invention.
Fig. 3 grid schematic diagrames for needed for being calculated the present invention.
Fig. 4 is tail end point relaxation path profile of the present invention.
Embodiment
Further to illustrate the present invention to reach the technological means and effect that predetermined goal of the invention is taken, below in conjunction with
Accompanying drawing and preferred embodiment, to according to its embodiment proposed by the present invention, structure, feature and its effect, describing in detail
As after.
As shown in figure 1, a kind of humming melody recognition methods based on dynamic time warp algorithm, comprises the following steps:
(1) voice messaging is pre-processed.The voice messaging of humming is subjected to denoising, the operation such as preemphasis and adding window framing is carried
The quality of high voice signal, making the HFS of signal becomes flat, and voice signal is integrally steady, is easy to follow-up analysis;
(2) feature extraction.End-point detection is carried out to the voice signal treated by above-mentioned steps and MFCC coefficients are extracted
Deng operation, the parameter extraction for representing voice signal substantive characteristics can be come out, obtain the characteristic sequence of humming input melody;
(3) the quick screening based on melody characteristicses.This method is used as feature melody according to half beat of song.Basis first
The pitch of song, which rises and falls, calculates the descant and contrabass half beat of a song.With melody to be measured half in delete target music libraries
Beat differs larger song.By this step, a part of song is quickly deleted, melody identification process is improved;
(4) the quick screening based on editing distance.The characteristic sequence for inputting melody will be hummed first poor according to pitch, conversion
Pass through into character string sequence, with target music libraries and remaining song progress calculating editor is screened based on melody characteristicses Rapid matching
Distance.Reject the song that distance differs larger;
(5) the accurate identification based on improved dynamic time warp algorithm.In humming identification, it will simply can not groan
Sing input melody characteristic sequence be compared with target template sequence because humming person there may be in humming it is various
Humming mistake, such as addition note deletes note etc., therefore it is particularly significant to carry out time planning to template.Dynamic time warp
Algorithm is by constantly calculating the distance of two vectors to find optimal coupling path, so two obtained vectors are cumulative distances
Minimum regular integer, it ensure that there are maximum acoustics similar features between them.But dynamic time warp algorithm is deposited
It is slow in retrieval time, the problems such as required memory space is big, therefore this method relaxes from end points regarding to the issue above, across sentence retrieval,
Several aspects such as cost function are improved to dynamic time warp algorithm.Realized by the dynamic time warp algorithm after improvement
The accurate identification of humming melody.
Quick screening principle of the above-mentioned steps (4) based on editing distance is as follows:
The characteristic sequence for inputting melody will be hummed first poor according to pitch, be converted into including the word of (E, U, X, S, D, B, T)
Pass through in symbol string sequence, with target musical database and quickly screen remaining song calculating editing distance based on melody characteristicses.Compile
It is a kind of method based on Dynamic Programming Idea to collect distance, and its principle is characteristic sequence x and characteristic sequence y editing distance, fixed
Justice is the minimum basic operation number needed for transforming to y from x, and basic operation here includes:
(a) character in x is replaced to be replaced by correspondence character string in y;
(b) character is inserted into x in insertion y, x length plus 1;
(c) character in x is deleted, is that x length subtracts 1.
Specific formula for calculation is as follows:
ED (a (i), b (j))=ED (a, b) if x=y
ED (a (i), b (j))=min (ED (a (i), b (j))+2, ED (a (i-1), b (j))+1, ED (a (i), b (j-1))+
1)
if x≠y
Wherein a, b are two feature strings, and ED examinations calculate a, and b subtracts the matrix of cumulative distance.
Accurate recognition principle of the above-mentioned steps (5) based on improved dynamic time warp algorithm is as follows:
Dynamic time warp algorithm is by constantly calculating the distance of two vectors to find optimal coupling path, so obtaining
To two vectors be the minimum regular integer of cumulative distance, it ensure that there are maximum acoustics similar features between them.It is dynamic
The state time, flexible algorithm principle was as shown in Figure 2.But because many actual mesh points are to be not required in actual matching process
Reach, thus Fig. 2 rhombus be a kind of limitation of path mode, the matching distance corresponding to lattice point so outside the rhombus is
It need not calculate, in addition also It is not necessary to preserve corresponding data, it is possible to reduce required memory space.Colleague, power is entered
One step reduces required amount of calculation, and the matching primitives needed for each lattice point have only used three grids of previous column, such as Fig. 3 institutes
Show.It is assumed here that there is two characteristic sequences to carry out dynamic time warp algorithm calculating, characteristic sequence to be matched is
X={ x1, x2, xn, template characteristic sequence is y={ y1, y2, yn, d is cost function, specific to calculate public
Formula is as follows:
D (i, j)=min (D (i-2, j-1)+di-2, j-1, D (i-1, j-1)+di-1,j-1, D (i-1, j-2)+di-1, j-2)
D is cost Korean style in above-mentioned formula, and specific formula for calculation is as follows:
di-1, j-1=abs (xi-yj)
di-2,j-1=abs (xi-1+xi-yj)+c1
di-1, j-2=abs (yj-1+yj-xi)+c2
Wherein, D is the matrix of accumulation distance between calculating X and Y, c1, c2For balance factor, it is therefore an objective to which balance insertion is deleted
The cost brought except note.
After two templates are matched, i.e., the minimum value from last column in distance accumulation matrix is recalled, from
The minimum value is once found forward accumulated each time before 3 points of minimum value, when returning to starting match point, then produce
One best match route.And the minimum value of last column of matrix is matched, it is exactly the distance of two templates.
Traditional dynamic time warp algorithm is all voice to be carried out head to head, matching of the tail to tail, but user hums
Melody there may be the situation of note insertion or missing, therefore clear and definite tail point can not be determined.This paper presents tail end
The method of point relaxation, concrete principle is as follows:
If P is that head-end is unknown, the Characteristic Number that N includes for humming input melody characteristicses sequence, w1, w2Respectively in length
Allow to insert note in the case of spending by N and delete the number of errors of note.In the case where head-end is determined, tail point area
Between be { (P+N-w2), (P+N+w1), increase path restrictive condition, calculate (P+N-w2) and (P+N+w1) two between position special
Levy between vector distance, selection minimum value and record, tail end point relaxation path such as Fig. 4.
The dynamic time warp algorithm for being conventionally used to humming identification only considered the feature of pitch, and have ignored the spy of the duration of a sound
Levy.This method is in order to improve the discrimination of humming identification, while being it is possible that humming section in view of humming person humming song
Play when the song rhythm in java standard library is fast or slow, the duration of a sound is introduced to the calculating of cost function.
If the characteristic sequence of humming input melody is X={ (tx1, ty1)(tx2, ty2)·(txm, tym), target database
In song features sequence be Y={ (rx1, ry1)(rx2, ry2)·(rxn, ryn), txi, rxjFor corresponding pitch difference sequence,
tyi, ryjCompare sequence for the corresponding duration of a sound.Improved cost function calculation is as follows:
di-1, j-1=u*abs (txi-rxj)+(1-u)*abs(tyi-ryj)*km
di-2, j-1=u*abs (txi-1+txi-rxj)+(1-u)*abs((tyi*tyi-1)/(1+tyi-1)-ryj)*km+c3
di-1, j-2=u*abs (rxj-1+rxj–txi)+(1-u)*abs((ryj*ryj-1)/(1+ryj-1)-ryi)*km+c4
Wherein c3, c4For balance factor, u is the weights introduced.This method finds that pitch parameters are more defined than otonaga features
Really, so setting u>0.5.Km be average pitch difference with the average duration of a sound than ratio.
Embodiments described above is only that the preferred embodiment of the present invention is described, not to the present invention's
Scope is defined, under the premise of design spirit of the present invention is not departed from, and this area ordinary skill technical staff is to the present invention program
In the various modifications made and improvement, the protection domain that claims of the present invention determination all should be fallen into.
Claims (3)
1. a kind of humming melody recognition methods based on dynamic time warp algorithm, it is characterised in that comprise the following steps:
(1) voice messaging is pre-processed.The voice messaging of humming is subjected to denoising, the operation such as preemphasis and adding window framing improves language
The quality of message number, making the HFS of signal becomes flat, and voice signal is integrally steady, is easy to follow-up analysis;
(2) feature extraction.The behaviour such as end-point detection and the extraction of MFCC coefficients is carried out to the voice signal treated by above-mentioned steps
Make, the parameter extraction for representing voice signal substantive characteristics can be come out, obtain the characteristic sequence of humming input melody;
(3) the quick screening based on melody characteristicses.This method is used as feature melody according to half beat of song.First according to song
Pitch rise and fall and calculate the descant and contrabass half beat of a first song.With the beat of melody half to be measured in delete target music libraries
The larger song of difference.By this step, a part of song is quickly deleted, melody identification process is improved;
(4) the quick screening based on editing distance.The characteristic sequence for inputting melody will be hummed first poor according to pitch, be converted into word
Accord with pass through in string sequence, with target music libraries based on melody characteristicses Rapid matching screen remaining song calculate editor away from
From.Reject the song that distance differs larger;
(5) the accurate identification based on improved dynamic time warp algorithm.In humming identification, it will simply can not hum defeated
The characteristic sequence and target template sequence for entering melody are compared, because humming person there may be various groan in humming
Mistake is sung, such as addition note, delete note etc., therefore it is particularly significant to carry out time planning to template.Dynamic time warp algorithm
By constantly calculating the distance of two vectors to find optimal coupling path, so two obtained vectors are that cumulative distance is minimum
Regular integer, it ensure that there are maximum acoustics similar features between them.But there is inspection in dynamic time warp algorithm
The rope time is slow, the problems such as required memory space is big, therefore this method relaxes from end points regarding to the issue above, across sentence retrieval, cost
Several aspects such as function are improved to dynamic time warp algorithm.Realized and groaned by the dynamic time warp algorithm after improvement
Sing the accurate identification of melody.
2. a kind of humming melody recognition methods based on dynamic time warp algorithm according to claim 1, its feature exists
The rapid screening method based on editing distance in the step (4) includes following principle:
The characteristic sequence that (2a) will hum input melody first is poor according to pitch, is converted into including the word of (E, U, X, S, D, B, T)
Pass through in symbol string sequence, with target musical database and quickly screen remaining song calculating editing distance based on melody characteristicses.Compile
It is a kind of method based on Dynamic Programming Idea to collect distance, and its principle is characteristic sequence x and characteristic sequence y editing distance, fixed
Justice is the minimum basic operation number needed for transforming to y from x, and basic operation here includes:
(a) character in x is replaced to be replaced by correspondence character string in y;
(b) character is inserted into x in insertion y, x length plus 1;
(c) character in x is deleted, is that x length subtracts 1.
(2b) specific formula for calculation is as follows:
ED (a (i), b (j))=ED (a, b) if x=y
ED (a (i), b (j))=min (ED (a (i), b (j))+2, ED (a (i-1), b (j))+1, ED (a (i), b (j-1))+1)
if x≠y
Wherein a, b are two feature strings, and ED examinations calculate a, and b subtracts the matrix of cumulative distance.
3. a kind of humming melody recognition methods based on dynamic time warp algorithm according to claim 1, its feature exists
The accurate identification based on improved dynamic time warp algorithm in the step (5) includes following principle steps:
(3a) traditional dynamic time warp algorithm is all voice to be carried out head to head, matching of the tail to tail, but user hums
Melody there may be the situation of note insertion or missing, therefore clear and definite tail point can not be determined.This paper presents tail end
The method of point relaxation, concrete principle is as follows:
If P is that head-end is unknown, the Characteristic Number that N includes for humming input melody characteristicses sequence, w1, w2It is N respectively in length
In the case of allow insert note and delete note number of errors.In the case where head-end is determined, tail point interval is
{(P+N-w2), (P+N+w1), increase path restrictive condition, calculate (P+N-w2) and (P+N+w1) two features between position to
Span between selection from minimum value and recording, tail end point relaxation path such as Fig. 4.
The dynamic time warp algorithm that (3b) is conventionally used to humming identification only considered the feature of pitch, and have ignored the spy of the duration of a sound
Levy.This method is in order to improve the discrimination of humming identification, while being it is possible that humming section in view of humming person humming song
Play when the song rhythm in java standard library is fast or slow, the duration of a sound is introduced to the calculating of cost function.
If the characteristic sequence of humming input melody is X={ (tx1, ty1)(tx2, ty2)…(txm, tym), in target database
Song features sequence is Y={ (rx1, ry1)(rx2, ry2)…(rxn, ryn), txi, rxjFor corresponding pitch difference sequence, tyi,
ryjCompare sequence for the corresponding duration of a sound.Improved cost function calculation is as follows:
dI-1, j-1=u*abs (txi-rxj)+(1-u)*abs(tyi-ryj)*km
dI-2, j-1=u*abs (txi-1+txi-rxj)+(1-u)*abs((tyi*tyi-1)/(1+tyi-1)-ryj)*km+c3
dI-1, j-2=u*abs (rxj-1+rxj–txi)+(1-u)*abs((ryj*ryj-1)/(1+ryj-1)-ryi)
*km+c4
Wherein c3, c4For balance factor, u is the weights introduced.This method finds that pitch parameters are more more accurate than otonaga features, institute
To set u>0.5.Km be average pitch difference with the average duration of a sound than ratio.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710648569.2A CN107256710A (en) | 2017-08-01 | 2017-08-01 | A kind of humming melody recognition methods based on dynamic time warp algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710648569.2A CN107256710A (en) | 2017-08-01 | 2017-08-01 | A kind of humming melody recognition methods based on dynamic time warp algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107256710A true CN107256710A (en) | 2017-10-17 |
Family
ID=60025472
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710648569.2A Pending CN107256710A (en) | 2017-08-01 | 2017-08-01 | A kind of humming melody recognition methods based on dynamic time warp algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107256710A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108320750A (en) * | 2018-01-23 | 2018-07-24 | 东南大学—无锡集成电路技术研究所 | A kind of implementation method based on modified dynamic time warping speech recognition algorithm |
CN108428441A (en) * | 2018-02-09 | 2018-08-21 | 咪咕音乐有限公司 | Multimedia file producting method, electronic equipment and storage medium |
CN108735231A (en) * | 2018-04-27 | 2018-11-02 | 大连民族大学 | Theme pitch sequence method of estimation |
CN111368129A (en) * | 2018-12-25 | 2020-07-03 | 天津大学青岛海洋技术研究院 | Humming retrieval method based on deep neural network |
CN113053337A (en) * | 2021-03-26 | 2021-06-29 | 北京儒博科技有限公司 | Intonation evaluation method, intonation evaluation device, intonation evaluation equipment and storage medium |
CN113377994A (en) * | 2021-07-08 | 2021-09-10 | 哈尔滨理工大学 | Humming retrieval method based on melody feature clustering and optimization |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1197952B1 (en) * | 2000-10-18 | 2009-12-02 | Thales | Coding method of the prosody for a very low bit rate speech encoder |
CN102053998A (en) * | 2009-11-04 | 2011-05-11 | 周明全 | Method and system device for retrieving songs based on voice modes |
KR20130077064A (en) * | 2011-12-29 | 2013-07-09 | 전자부품연구원 | Fast music information retrieval system based on query by humming and method thereof |
CN103366784A (en) * | 2013-07-16 | 2013-10-23 | 湖南大学 | Multimedia playing method and device with function of voice controlling and humming searching |
-
2017
- 2017-08-01 CN CN201710648569.2A patent/CN107256710A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1197952B1 (en) * | 2000-10-18 | 2009-12-02 | Thales | Coding method of the prosody for a very low bit rate speech encoder |
CN102053998A (en) * | 2009-11-04 | 2011-05-11 | 周明全 | Method and system device for retrieving songs based on voice modes |
KR20130077064A (en) * | 2011-12-29 | 2013-07-09 | 전자부품연구원 | Fast music information retrieval system based on query by humming and method thereof |
CN103366784A (en) * | 2013-07-16 | 2013-10-23 | 湖南大学 | Multimedia playing method and device with function of voice controlling and humming searching |
Non-Patent Citations (1)
Title |
---|
JIN LI等: "Improved Dynamic Time Warping Algorithm the research and application of Query by Humming", 《2010 SIXTH INTERNATIONAL CONFERENCE ON NATURAL COMPUTATION》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108320750A (en) * | 2018-01-23 | 2018-07-24 | 东南大学—无锡集成电路技术研究所 | A kind of implementation method based on modified dynamic time warping speech recognition algorithm |
CN108428441A (en) * | 2018-02-09 | 2018-08-21 | 咪咕音乐有限公司 | Multimedia file producting method, electronic equipment and storage medium |
CN108428441B (en) * | 2018-02-09 | 2021-08-06 | 咪咕音乐有限公司 | Multimedia file generation method, electronic device and storage medium |
CN108735231A (en) * | 2018-04-27 | 2018-11-02 | 大连民族大学 | Theme pitch sequence method of estimation |
CN108735231B (en) * | 2018-04-27 | 2021-11-12 | 大连民族大学 | Method for estimating pitch sequence of main melody |
CN111368129A (en) * | 2018-12-25 | 2020-07-03 | 天津大学青岛海洋技术研究院 | Humming retrieval method based on deep neural network |
CN113053337A (en) * | 2021-03-26 | 2021-06-29 | 北京儒博科技有限公司 | Intonation evaluation method, intonation evaluation device, intonation evaluation equipment and storage medium |
CN113377994A (en) * | 2021-07-08 | 2021-09-10 | 哈尔滨理工大学 | Humming retrieval method based on melody feature clustering and optimization |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107256710A (en) | A kind of humming melody recognition methods based on dynamic time warp algorithm | |
US8013229B2 (en) | Automatic creation of thumbnails for music videos | |
US7342167B2 (en) | Apparatus and method for generating an encoded rhythmic pattern | |
US7064262B2 (en) | Method for converting a music signal into a note-based description and for referencing a music signal in a data bank | |
Kroher et al. | Automatic transcription of flamenco singing from polyphonic music recordings | |
Vogl et al. | Recurrent Neural Networks for Drum Transcription. | |
JP2010054802A (en) | Unit rhythm extraction method from musical acoustic signal, musical piece structure estimation method using this method, and replacing method of percussion instrument pattern in musical acoustic signal | |
KR101520621B1 (en) | / Method and apparatus for query by singing/huming | |
Yoshii et al. | Automatic Drum Sound Description for Real-World Music Using Template Adaptation and Matching Methods. | |
JP2012234202A (en) | Rhythm structure extraction method, method for determining analogous relation between items of plural audio signal, and program | |
Lee et al. | Automatic Chord Recognition from Audio Using a HMM with Supervised Learning. | |
CN109979488A (en) | Voice based on stress analysis turns music notation system | |
WO2017154928A1 (en) | Audio signal processing method and audio signal processing device | |
Eggink et al. | Instrument recognition in accompanied sonatas and concertos | |
JP2002116754A (en) | Tempo extraction device, tempo extraction method, tempo extraction program and recording medium | |
Wang et al. | Robust and efficient joint alignment of multiple musical performances | |
CN110399522A (en) | A kind of music singing search method and device based on LSTM and layering and matching | |
Atli et al. | Audio feature extraction for exploring Turkish makam music | |
CN113192471B (en) | Musical main melody track recognition method based on neural network | |
JP3934556B2 (en) | Method and apparatus for extracting signal identifier, method and apparatus for creating database from signal identifier, and method and apparatus for referring to search time domain signal | |
CN112634841B (en) | Guitar music automatic generation method based on voice recognition | |
JP4765971B2 (en) | Mixed model generation apparatus, sound processing apparatus, and program | |
JP6056799B2 (en) | Program, information processing apparatus, and data generation method | |
JP4305509B2 (en) | Voice processing apparatus and program | |
JP2017161572A (en) | Sound signal processing method and sound signal processing device |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171017 |
|
WD01 | Invention patent application deemed withdrawn after publication |