CN106205572B - Sequence of notes generation method and device - Google Patents
Sequence of notes generation method and device Download PDFInfo
- Publication number
- CN106205572B CN106205572B CN201610488830.2A CN201610488830A CN106205572B CN 106205572 B CN106205572 B CN 106205572B CN 201610488830 A CN201610488830 A CN 201610488830A CN 106205572 B CN106205572 B CN 106205572B
- Authority
- CN
- China
- Prior art keywords
- note
- sequence
- vector
- notes
- note vector
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000013598 vector Substances 0.000 claims abstract description 201
- 238000012549 training Methods 0.000 claims abstract description 34
- 230000011218 segmentation Effects 0.000 claims description 5
- 230000001755 vocal effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 23
- 230000008569 process Effects 0.000 description 8
- 241000288673 Chiroptera Species 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- 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
- G10H1/00—Details of electrophonic musical instruments
- G10H1/0008—Associated control or indicating means
- G10H1/0025—Automatic or semi-automatic music composition, e.g. producing random music, applying rules from music theory or modifying a musical piece
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10G—REPRESENTATION OF MUSIC; RECORDING MUSIC IN NOTATION FORM; ACCESSORIES FOR MUSIC OR MUSICAL INSTRUMENTS NOT OTHERWISE PROVIDED FOR, e.g. SUPPORTS
- G10G3/00—Recording music in notation form, e.g. recording the mechanical operation of a musical instrument
- G10G3/04—Recording music in notation form, e.g. recording the mechanical operation of a musical instrument using electrical means
-
- 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
- G10H1/00—Details of electrophonic musical instruments
- G10H1/0033—Recording/reproducing or transmission of music for electrophonic musical instruments
- G10H1/0041—Recording/reproducing or transmission of music for electrophonic musical instruments in coded form
-
- 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
-
- 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
- G10H2210/101—Music Composition or musical creation; Tools or processes therefor
- G10H2210/111—Automatic composing, i.e. using predefined musical rules
- G10H2210/115—Automatic composing, i.e. using predefined musical rules using a random process to generate a musical note, phrase, sequence or structure
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Auxiliary Devices For Music (AREA)
Abstract
The invention discloses a kind of sequence of notes generation method and devices, belong to intelligent vocal music field.Method includes: to obtain first note sequence, and first note sequence includes at least one note vector；The probability value of each note vector at least one note vector is determined based on first note sequence and sequence of notes model, the probability value of i-th of note vector is used to indicate the probability that i-th of note vector is next note vector after first note sequence at least one note vector, i is more than or equal to 1, sequence of notes model includes at least one function, and the parameter at least one function is obtained according to sequence of notes sample training；Any note vector that probability value is greater than preset threshold is determined as target note vector；The addition of target note vector is generated into the second sequence of notes after first note sequence.The invention enables the users for not having music theory knowledge and inspiration scarcity can also obtain original melody.
Description
Technical field
The present invention relates to intelligent vocal music field, in particular to a kind of sequence of notes generation method and device.
Background technique
With the development of music, demand of the user to original music is increasing, i.e., user is intended not only to obtain professional sound
The music of happy creator's creation, it is also desirable to which oneself being capable of art music.However, requirement of the musical composition to user is very high, to
Musical composition is completed not only to need to possess music theory knowledge, have certain musical quality, it is also necessary to certain creation inspiration, and one
As user often lack music theory knowledge, and creation inspiration is also deficienter, and it is very difficult to carry out musical composition.
Summary of the invention
In order to solve problems in the prior art, the embodiment of the invention provides a kind of sequence of notes generation method and devices.
The technical solution is as follows:
On the one hand, a kind of sequence of notes generation method is provided, which comprises
First note sequence is obtained, the first note sequence includes at least one note vector, and the note vector is used
In the pitch value and note duration of instruction note；
Based on the first note sequence and sequence of notes model determine at least one note vector each note to
The probability value of amount, at least one described note vector the probability value of i-th of note vector be used to indicate i-th of note to
Amount is the probability of next note vector after the first note sequence, and the i is more than or equal to 1, the sequence of notes mould
Type includes at least one function, and the parameter at least one described function is obtained according to sequence of notes sample training；
Any note vector that the probability value is greater than preset threshold is determined as the target note vector；
Target note vector addition is generated into the second sequence of notes after the first note sequence.
On the other hand, a kind of sequence of notes generating means are provided, described device includes:
Module is obtained, for obtaining first note sequence, the first note sequence includes at least one note vector, institute
State pitch value and note duration that note vector is used to indicate note；
Probability determination module, the first note sequence and sequence of notes model for being obtained based on the acquisition module
Determine the probability value of each note vector at least one note vector, i-th of note at least one described note vector
It is next note vector after the first note sequence that the probability value of vector, which is used to indicate i-th of note vector,
Probability, the i are more than or equal to 1, and the sequence of notes model includes at least one function, the parameter at least one described function
It is obtained according to sequence of notes sample training；
Vector determining module, any note vector for the probability value to be greater than preset threshold are determined as the target
Note vector；
Sequence generating module, the target note vector addition for determining the vector determining module is described the
The second sequence of notes is generated after one sequence of notes.
Technical solution provided in an embodiment of the present invention has the benefit that
Target note vector is determined by the first note sequence and sequence of notes model, and the target note vector is added
It adds to after the first note sequence, forms the second sequence of notes, allow the present invention according only to the first sound obtained in advance
The second sequence of notes can be obtained in symbol sequence, without the creation of user, so that not having music theory knowledge and inspiration
Deficient user can also obtain original melody.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of flow chart of sequence of notes generation method provided in an embodiment of the present invention.
Fig. 2 is the flow chart of another sequence of notes generation method provided in an embodiment of the present invention.
Fig. 3 is a kind of block diagram of sequence of notes generating means 300 provided in an embodiment of the present invention.
Fig. 4 is a kind of block diagram of sequence of notes generating means 400 provided in an embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Fig. 1 is a kind of flow chart of sequence of notes generation method shown according to an exemplary embodiment, as shown in Figure 1,
The sequence of notes generation method can be applied in server or terminal, the sequence of notes generation method the following steps are included:
Step 110 obtains first note sequence, which includes at least one note vector, the note to
Amount is used to indicate the pitch value and note duration of note.
Step 120 determines each at least one note vector based on the first note sequence and sequence of notes model
The probability value of note vector, the probability value of i-th of note vector is used to indicate i-th of note at least one note vector
Vector is the probability of next note vector after the first note sequence, which is more than or equal to 1, the sequence of notes model packet
At least one function is included, the parameter at least one function is obtained according to sequence of notes sample training.
Any note vector that the probability value is greater than preset threshold is determined as the target note vector by step 130.
Target note vector addition is generated the second sequence of notes after the first note sequence by step 140.
In conclusion sequence of notes generation method provided in this embodiment, passes through the first note sequence and sequence of notes
Model determines target note vector, and after the target note vector is added to the first note sequence, forms the second note
Sequence allows the present invention that the second sequence of notes can be obtained according only to the first note sequence obtained in advance, without
The creation of user, so that the user for not having music theory knowledge and inspiration scarcity can also obtain original melody.
Optionally, each at least one note vector should be determined based on the first note sequence and sequence of notes model
Before the probability value of note vector, this method further include:
Judge whether the number for the note vector for including in the first note sequence is less than first threshold；
Should be determined at least one note vector based on the first note sequence and sequence of notes model each note to
The probability value of amount includes:
When the number for the note vector for including in the first note sequence is less than the first threshold, it is based on the first note
Sequence and the sequence of notes model determine the probability value of each note vector at least one note vector.
Optionally, each at least one note vector should be determined based on the first note sequence and sequence of notes model
Before the probability value of note vector, this method further include:
At least one specified sequence of notes is obtained, which includes at least one note vector；
By this, the specified note sequences segmentation of each of at least one specified sequence of notes accords with sequence at least one consonant
Column, the note vector number for each including in the sub- sequence of notes are less than second threshold；
Based at least one, the sub- sequence of notes training obtains the sequence of notes model.
All the above alternatives can form alternative embodiment of the invention using any combination, herein no longer
It repeats one by one.
Fig. 2 is a kind of flow chart of sequence of notes generation method shown according to an exemplary embodiment, as shown in Fig. 2,
The sequence of notes generation method can be applied in server or terminal, the present embodiment only in this way be applied to server in be
Example be illustrated, the sequence of notes generation method the following steps are included:
Step 210, server obtain first note sequence, which includes at least one note vector, should
Note vector is used to indicate the pitch value and note duration of note.
It should be noted that above-mentioned note vector is used to indicate note, wherein when a note includes pitch value and note
Value, moreover, a note vector is generally used for one note of instruction.
Above-mentioned pitch value is used to reflect the tone height of note, and in one embodiment of the invention, which can be with
It is divided into a times bass, bass, middle pitch, high pitch and five area Ge Yin of high pitch again, wherein each sound area can be divided into seven scales.Certainly,
In practical applications, which can also include other sound areas such as a times times bass, times times high pitch, alternatively, the pitch value can be with
Including bass again, bass, middle pitch, high pitch and again any of high pitch or appoint multiple combination, the present invention does not do this specifically
It limits.In addition, each sound area is further divided into five scales, if plumage or 12 scales etc. are levied in palace quotient angle, the present invention is to this
It is not specifically limited.
Above-mentioned note duration is used to reflect the relative duration of note, which may include whole note, two points
Note, crotchet, quaver, semiquaver and demisemiquaver, wherein the duration of whole note expression note
For four bats, minim indicates that the duration of note is two bats, and crotchet indicates that the duration of note claps for one, eight points
The duration of note expression note is half bat, and semiquaver indicates that the duration of note is that a quarter is clapped, 30 points
Note indicates that the duration of note is 1/8th bats.
It should be noted that in practical applications, above-mentioned note can also include rest, for rest,
Pitch value can be considered as 0, and its note duration may include above-mentioned whole note, minim, crotchet, quaver, ten
Six dieresis and demisemiquaver etc., the present invention is not specifically limited in this embodiment.For example, " 0000 " is semibreverest, sound
Tone pitch is 0, and note duration is whole note.
For example, a certain note can beIts pitch value indicated is the 5th scale of treble, the note indicated
Duration is whole note, i.e. the duration of note is four bats.
As described above, note vector is used to indicate note, in practical application, note vector may include it is multiple in sequence
The number of arrangement, for example, a certain note vector can be [0,1,0,0 ... ..., 0], wherein include 210 in the note vector
Number.The method for generating the corresponding note vector of notes the present invention provides two kinds, below the present invention will to both methods into
Row brief description.
First method is that each note sets corresponding note vector by technical staff.The present invention is with pitch value point
For five areas Ge Yin, each sound divides into 7 scales and note duration includes whole note, minim, crotchet, eight points
It is that each note sets corresponding note vector to technical staff for note, semiquaver and demisemiquaver
Technical process is illustrated.
Specifically, it is divided into five areas Ge Yin in pitch value, each sound divides into 7 scales and note duration includes whole tone
In the case where symbol, minim, crotchet, quaver, semiquaver and demisemiquaver, the quantity of note is in total
There are 210.Then technical staff can set each note vector include 210 numbers, i.e., note vector include number
Number be equal to note total number, and each note vector include 210 numbers in only one number be " 1 ", other number
Word is " 0 ", and position of the different note vectors " 1 " in 210 numbers is different.For example,Corresponding note vector
Can be [1,0,0,0 ... ..., 0],Corresponding note vector can be [0,1,0,0 ... ..., 0].
Technical staff is the method that each note sets corresponding note vector, on the one hand can make the expression number of note
Huas facilitates carry out subsequent processing, and on the other hand, the corresponding note vector of acquisition note does not need server and carries out at calculating
Reason, it is possible to reduce the operand of server.
Second method calculates the corresponding note vector of each note by sequence of notes model training.Similarly,
The note vector obtained with this method also includes multiple numbers arranged in sequence, and, include in each note vector
Number number can be set by technical staff.
Specifically, the available multiple sequence of notes of the present invention are as sequence of notes model training sample to sequence of notes mould
Pair type is trained, and in one embodiment of the invention, skip-gram model can be used in sequence of notes model training, i.e.,
In any one sequence of notes x1..., xT, make objective functionIt maximizes, wherein c mono-
The note number for including in a sequence of notes model training sample.Wherein, p (xt+j|xj) softmax function representation can be used,
I.e.Wherein, v is note vector, and X is the total quantity of note (for example, being divided into five in pitch value
The area Ge Yin, each sound divide into 7 scales and note duration include whole note, minim, crotchet, quaver,
In the case where semiquaver and demisemiquaver, X 210).
The method for calculating the corresponding note vector of each note by sequence of notes model training, on the one hand can also be with
The expression mathematicization for making note facilitates carry out subsequent processing, on the other hand, includes in the note vector generated in this way
Number number can be less than note total number, the meter in technical process so as to reduce subsequent step 220 to 250
Calculation amount, for example, in one embodiment of the invention, can only be wrapped using the calculated note vector of sequence of notes model training
Containing 100 numbers.
As described above, needing to obtain multiple sequence of notes in the method for generating the corresponding note vector of note at second
As sequence of notes model training sample.Specifically, the present invention can first obtain multiple music score, include multiple sounds in each music score
Symbol, then carries out participle operation to each music score, neglects other symbols in addition to note for including in music score, such as trifle
Line, liaison line etc., to form sequence of notes, multiple notes that multiple music score is formed after participle operation are can be used in server
Sequence is as sequence of notes model training sample.
In addition it is also necessary to illustrate, the number for the note vector for including in above-mentioned first note sequence can be one
Or it is multiple.When the number of note vector in the first note sequence is one, one in the first note sequence
Note vector can be the note vector that server determines at random, or server operates determining according to the user's choice
Note vector, the present invention are not specifically limited in this embodiment.When the number of note vector in the first note sequence is more than one,
The first note sequence can be the sequence of notes that server determines at random, or server operates according to the user's choice
Determining sequence of notes can also be the sequence of notes that server is generated according to subsequent step 220 to 250, and the present invention is to this
It is not specifically limited.
Whether step 220, server judge the number for the note vector for including in the first note sequence less than the first threshold
Value, if being less than, thens follow the steps 230, if being not less than, terminates process.
Sequence of notes can be generated in sequence of notes generation method provided by the invention, at least one note in the sequence of notes
At least one note indicated by vector can form a kind of art, pleasant sound, that is, form melody.Actually answering
In, melody generally has certain length, therefore, in the present invention, the note for including in the sequence of notes that server generates
The number of vector should also be as being greater than a certain threshold value, i.e. first threshold, and otherwise, at least one note vector is signified in the sequence of notes
The number of at least one note shown will be not enough to generate melody.Therefore, if the note vector for including in the first note sequence
Number be no less than first threshold, then illustrate in the first note sequence at least one indicated by least one note vector
The number of note has been enough to form melody, and the present invention can terminate process at this time, if the sound for including in the first note sequence
The number for according with vector is less than first threshold, then illustrates in the first note sequence at least one indicated by least one note vector
The number of a note is not enough to be formed melody, then the present invention can execute the technical process of step 230.
Step 230, server are determined at least one note vector based on the first note sequence and sequence of notes model
The probability value of each note vector, at least one note vector the probability value of i-th of note vector be used to indicate this i-th
A note vector is the probability of next note vector after the first note sequence, which is more than or equal to 1, the sequence of notes
Model includes at least one function, and the parameter at least one function is obtained according to sequence of notes sample training.
In the present invention, server can be using the first note sequence as the input value of sequence of notes model, to utilize
The sequence of notes model calculates the probability value of each note vector at least one note vector.At least one above-mentioned note to
Amount can be all note vectors, for example, being divided into five areas Ge Yin in pitch value, each sound divides into 7 scales and note
In the case that duration includes whole note, minim, crotchet, quaver, semiquaver and demisemiquaver, institute
Some note vectors are the corresponding note vector of each note in 210 different notes, certainly, at least one above-mentioned note
Vector can be a part of note vector in all note vectors, and the present invention is not specifically limited in this embodiment.
It should be noted that the probability value of above-mentioned note vector is used to indicate note vector as after the first note sequence
Next note vector probability, the biggish note vector of probability value be the first note sequence after next note to
The probability of amount is also larger, i.e., the biggish note vector of probability value is placed on the first note sequence later and the second note of generation
Sequence is that the probability of melody is bigger.
Exemplary, above-mentioned sequence of notes model can be recurrent neural networks model, which includes input layer, hidden layer
And output layer, wherein input layer is basecoat, and output layer is most upper one layer, hidden layer between input layer and output layer,
" input layer " is used to receive the input value for the sequence of notes model, i.e., " first note sequence " described above is " implicit
Layer " output valve can by " input layer " input input value directly or indirectly obtain, sequence of notes model it is final defeated
Value is obtained by above-mentioned " output layer " out, i.e., the probability of each note vector at least one note vector described above
Value.
Wherein, input layer, hidden layer and output layer include at least one node, and the node in every layer passes through directed arc
It is directed toward a upper node layer, each node in hidden layer and output layer corresponds to a function, and the parameter of the function is to be directed toward to be somebody's turn to do
The corresponding weighted value of the directed arc of node, the input value of the function are to be directed toward the output valve of the lower level node of the node, the function
Output valve can be obtained based on the parameter and the input value.Wherein, the corresponding weighted value of directed arc, the i.e. parameter of the function can be with
It is obtained according to sequence of notes sample training.
Since the final purpose of the present invention is to generate melody, that is, sequence of notes is generated, therefore the present invention is available a large amount of
Sequence of notes sample, the present invention is before to the sequence of notes model training, although can not determine in the sequence of notes model
The parameter for including, still, the present invention is based on input value and output valve that the sequence of notes sample knows the sequence of notes model, bases
In a large amount of input value and output valve, the present invention may finally train to obtain the parameter for including in the sequence of notes model.
Exemplary, in one embodiment of the invention, the function for including in sequence of notes model provided by the invention can
With are as follows:
st=tanh (Uxt+Wst-1)
ot=soft max (Vst)
Wherein, xtFor t-th of note vector in first note sequence, stFor the output valve of t layers of hidden layer, otIt is defeated
The output of the output valve of layer out, i.e. model is the probability value of each note vector at least one note vector, and U, W, V is letter
Several parameters.
As described above, the sequence of notes model can be obtained according to sequence of notes sample training, above-mentioned sequence of notes sample
Acquisition modes and the multiple sequence of notes of acquisition described above it is similar as the mode of sequence of notes model training sample, no
It is to carry out participle operation to each music score with place, neglects other symbols in addition to note for including in music score, it is such as small
Nodel line, liaison line etc. also need each note in sequence of notes being converted to note vector, with final after forming sequence of notes
Obtain sequence of notes sample.
In the following, the present invention will carry out briefly the technical process for obtaining sequence of notes model according to sequence of notes sample training
Explanation.
Specifically, server obtains at least one specified sequence of notes, and by this at least one specified sequence of notes
For each specified note sequences segmentation at least one sub- sequence of notes, the note vector for each including in the sub- sequence of notes is a
Number is less than second threshold, and then server is based at least one sub- sequence of notes training and obtains the sequence of notes model.
It should be noted that above-mentioned " at least one specified sequence of notes " is " sequence of notes sample " described above.
The present invention is by server, based at least one, the sub- sequence of notes training obtains the sequence of notes model below
Technical process is briefly described.
Specifically, the present invention can be used recurrent neural networks model and be trained to sequence of notes model, i.e., for appointing
One sub- sequence of notes x1... ..., xT, make objective functionIt maximizes, wherein p
(xt|x1.. ..., xt)=g (ht), g is sigmoid function or tanh function, and htIt can be indicated with following formula:
Wherein, φ may be sigmoid function, be completed using the continuous adjusting parameter of gradient descent method to the note
The training of series model.
Any note vector that the probability value is greater than preset threshold is determined as target note vector by step 240, server.
In one embodiment of the invention, target note vector can be probability value at least one above-mentioned note vector
Maximum note vector.In practical application, the method that server can also use multinomial distribution to sample extracts maximum probability
Value, and the corresponding note vector of the maximum probability value is determined as target note vector.
Specifically, if p1=p (x=X1) ... ..., pk=p (x=Xk), wherein XiFor at least one above-mentioned note vector
I-th of note vector, p are the discrete multinomial distributions of x, and the random number u~U (0,1) for randomly selecting a 0-1, which is obeyed, uniformly to be divided
Cloth, by p1, p2... ... pkThe cumulative corresponding X of k no more than ukAs the next note sampled out.
For example, the probability value of each note vector is as shown in table 1 at least one note vector:
Table 1
| Note vector | X1 | X2 | X3 | X4 | X5 | …… | X210 |
| Probability value | 0.0001 | 0.0002 | 0.0003 | 0.0001 | 0.0001 | …… | 0.98 |
In table 1, the method that multinomial distribution sampling can be used in the present invention, by X210It is determined as target note vector.
Target note vector addition is generated the second note sequence by step 250, server after the first note sequence
Column.
In the present invention, after generating the second sequence of notes, the technology mistake of step 220 to step 250 can be repeated
Journey.
In conclusion sequence of notes generation method provided in this embodiment, passes through the first note sequence and sequence of notes
Model determines target note vector, and after the target note vector is added to the first note sequence, forms the second note
Sequence allows the present invention that the second sequence of notes can be obtained according only to the first note sequence obtained in advance, without
The creation of user, so that the user for not having music theory knowledge and inspiration scarcity can also obtain original melody.
Fig. 3 is a kind of block diagram of sequence of notes generating means 300 shown according to an exemplary embodiment.It, should referring to Fig. 3
Device includes obtaining module 310, probability determination module 320, vector determining module 330 and sequence generating module 340.
The acquisition module 310, for obtaining first note sequence, the first note sequence include at least one note to
Amount, the note vector are used to indicate the pitch value and note duration of note.
The probability determination module 320, the first note sequence and sequence of notes for being obtained based on the acquisition module 310
Model determines the probability value of each note vector at least one note vector, i-th of sound at least one note vector
It is the general of next note vector after the first note sequence that the probability value of symbol vector, which is used to indicate i-th of note vector,
Rate, the i are more than or equal to 1, which includes at least one function, and the parameter at least one function is according to note
Sequence samples training obtains.
The vector determining module 330, any note vector for the probability value to be greater than preset threshold are determined as the mesh
Mark with phonetic symbols accords with vector.
The sequence generating module 340, the target note vector for determining the vector determining module 330 are added at this
The second sequence of notes is generated after first note sequence.
In one embodiment of the invention, the sequence generating module 340 is also used to judge that the acquisition module 310 obtains
The first note sequence in include the number of note vector whether be less than first threshold.
The probability determination module 320, the number for being also used to the note vector for including in the first note sequence, which is less than, is somebody's turn to do
When first threshold, each note at least one note vector is determined based on the first note sequence and the sequence of notes model
The probability value of vector.
The acquisition module 310 is also used to obtain at least one specified sequence of notes, which includes at least one
A note vector.
Referring to fig. 4, in another embodiment of the present invention, another sequence of notes generating means 400 are additionally provided, it should
Example structure of the device based on above-mentioned Fig. 3 further includes training module 350.
The training module 350, for specified note sequences segmentation of each of at least one specified sequence of notes by this
At at least one sub- sequence of notes, the note vector number for including in each sub- sequence of notes is less than second threshold.
The training module 350 is also used to the sub- sequence of notes training based at least one and obtains the sequence of notes model.
In conclusion sequence of notes generating means provided in this embodiment, pass through the first note sequence and sequence of notes
Model determines target note vector, and after the target note vector is added to the first note sequence, forms the second note
Sequence allows the present invention that the second sequence of notes can be obtained according only to the first note sequence obtained in advance, without
The creation of user, so that the user for not having music theory knowledge and inspiration scarcity can also obtain original melody.
It should be understood that sequence of notes generating means provided by the above embodiment are in formation sequence, only with above-mentioned each
The division progress of functional module can according to need and for example, in practical application by above-mentioned function distribution by different function
Energy module is completed, i.e., the internal structure of device is divided into different functional modules, to complete whole described above or portion
Divide function.In addition, sequence of notes generating means provided by the above embodiment belong to sequence of notes generation method embodiment it is same
Design, specific implementation process are detailed in embodiment of the method, and which is not described herein again.
Those of ordinary skill in the art will appreciate that realizing that all or part of the steps of above-described embodiment can pass through hardware
It completes, relevant hardware can also be instructed to complete by program, the program can store in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only memory, disk or CD etc..
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (6)
1. a kind of sequence of notes generation method, which is characterized in that the described method includes:
First note sequence is obtained, the first note sequence includes at least one note vector, and the note vector is for referring to
Show the pitch value and note duration of note；
Each note vector at least one note vector is determined based on the first note sequence and sequence of notes model
Probability value, the probability value of i-th of note vector is used to indicate i-th of note vector and is at least one described note vector
The probability of next note vector after the first note sequence, the i are more than or equal to 1, the sequence of notes model packet
At least one function is included, the parameter at least one described function is obtained according to sequence of notes sample training；
Any note vector that the probability value is greater than preset threshold is determined as target note vector；
Target note vector addition is generated into the second sequence of notes after the first note sequence.
2. the method according to claim 1, wherein described be based on the first note sequence and sequence of notes mould
Type determines at least one note vector before the probability value of each note vector, the method also includes:
Judge whether the number for the note vector for including in the first note sequence is less than first threshold；
It is described based on the first note sequence and sequence of notes model determine at least one note vector each note to
The probability value of amount includes:
When the number for the note vector for including in the first note sequence is less than the first threshold, it is based on first sound
Symbol sequence and the sequence of notes model determine the probability value of each note vector at least one note vector.
3. the method according to claim 1, wherein described be based on the first note sequence and sequence of notes mould
Type determines at least one note vector before the probability value of each note vector, the method also includes:
At least one specified sequence of notes is obtained, the specified sequence of notes includes at least one note vector；
Will the specified note sequences segmentation of each of at least one described specified sequence of notes at least one sub- sequence of notes,
The note vector number for including in each sub- sequence of notes is less than second threshold；
The sequence of notes model is obtained based on the training of sub- sequence of notes described at least one.
4. a kind of sequence of notes generating means, which is characterized in that described device includes:
Module is obtained, for obtaining first note sequence, the first note sequence includes at least one note vector, the sound
Symbol vector is used to indicate the pitch value and note duration of note；
Probability determination module, the first note sequence and sequence of notes model for being obtained based on the acquisition module are determined
The probability value of each note vector at least one note vector, i-th of note vector at least one described note vector
Probability value be used to indicate i-th of note vector be the first note sequence after next note vector it is general
Rate, the i are more than or equal to 1, and the sequence of notes model includes at least one function, the parameter root at least one described function
It is obtained according to sequence of notes sample training；
Vector determining module, for by the probability value be greater than preset threshold any note vector be determined as target note to
Amount；
Sequence generating module, the target note vector for determining the vector determining module are added in first sound
It accords with sequence and generates the second sequence of notes later.
5. device according to claim 4, which is characterized in that the sequence generating module is also used to judge the acquisition
Whether the number for the note vector for including in the first note sequence that module obtains is less than first threshold；
The probability determination module, the number for being also used in the first note sequence note vector for including are less than described the
When one threshold value, each sound at least one note vector is determined based on the first note sequence and the sequence of notes model
Accord with the probability value of vector.
6. device according to claim 4, which is characterized in that it is specified to be also used to obtain at least one for the acquisition module
Sequence of notes, the specified sequence of notes include at least one note vector；
The sequence of notes generating means further include training module, the training module, at least one designated tone by described in
The specified note sequences segmentation of each of sequence is accorded at least one sub- sequence of notes, includes in each sub- sequence of notes
Note vector number be less than second threshold；
The training module is also used to obtain the sequence of notes model based at least one sub- sequence of notes training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610488830.2A CN106205572B (en) | 2016-06-28 | 2016-06-28 | Sequence of notes generation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610488830.2A CN106205572B (en) | 2016-06-28 | 2016-06-28 | Sequence of notes generation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106205572A CN106205572A (en) | 2016-12-07 |
| CN106205572B true CN106205572B (en) | 2019-09-20 |
Family
ID=57462084
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610488830.2A Expired - Fee Related CN106205572B (en) | 2016-06-28 | 2016-06-28 | Sequence of notes generation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106205572B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107123415B (en) * | 2017-05-04 | 2020-12-18 | 吴振国 | Automatic song editing method and system |
| CN113516961B (en) * | 2021-09-15 | 2021-12-03 | 腾讯科技（深圳）有限公司 | Note generation method, related device, storage medium and program product |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4662262A (en) * | 1985-03-08 | 1987-05-05 | Casio Computer Co., Ltd. | Electronic musical instrument having autoplay function |
| CN1049566A (en) * | 1990-03-21 | 1991-02-27 | 白亦民 | Intelligent electric musical instrument |
| CN1328679A (en) * | 1998-09-24 | 2001-12-26 | 马达尔萨尔公司 | Automatic music generating method and device |
| CN105118490A (en) * | 2015-07-20 | 2015-12-02 | 科大讯飞股份有限公司 | Polyphonic musical instrument note positioning method and polyphonic musical instrument note positioning device |
| CN105161087A (en) * | 2015-09-18 | 2015-12-16 | 努比亚技术有限公司 | Automatic harmony method, device, and terminal automatic harmony operation method |
| CN105632474A (en) * | 2014-11-20 | 2016-06-01 | 卡西欧计算机株式会社 | Automatic composition apparatus and method and storage medium |
| CN105632476A (en) * | 2014-11-20 | 2016-06-01 | 卡西欧计算机株式会社 | Automatic composition apparatus and method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6815600B2 (en) * | 2002-11-12 | 2004-11-09 | Alain Georges | Systems and methods for creating, modifying, interacting with and playing musical compositions |
-
2016
- 2016-06-28 CN CN201610488830.2A patent/CN106205572B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4662262A (en) * | 1985-03-08 | 1987-05-05 | Casio Computer Co., Ltd. | Electronic musical instrument having autoplay function |
| CN1049566A (en) * | 1990-03-21 | 1991-02-27 | 白亦民 | Intelligent electric musical instrument |
| CN1328679A (en) * | 1998-09-24 | 2001-12-26 | 马达尔萨尔公司 | Automatic music generating method and device |
| CN105632474A (en) * | 2014-11-20 | 2016-06-01 | 卡西欧计算机株式会社 | Automatic composition apparatus and method and storage medium |
| CN105632476A (en) * | 2014-11-20 | 2016-06-01 | 卡西欧计算机株式会社 | Automatic composition apparatus and method |
| CN105118490A (en) * | 2015-07-20 | 2015-12-02 | 科大讯飞股份有限公司 | Polyphonic musical instrument note positioning method and polyphonic musical instrument note positioning device |
| CN105161087A (en) * | 2015-09-18 | 2015-12-16 | 努比亚技术有限公司 | Automatic harmony method, device, and terminal automatic harmony operation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106205572A (en) | 2016-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Manaris et al. | Evolutionary music and the Zipf-Mandelbrot law: Developing fitness functions for pleasant music | |
| CN108920622A (en) | A kind of training method of intention assessment, training device and identification device | |
| Van Der Merwe et al. | Music generation with markov models | |
| CN109166564A (en) | For the method, apparatus and computer readable storage medium of lyrics text generation melody | |
| CN109346045A (en) | Counterpoint generation method and device based on long neural network in short-term | |
| CN106503034A (en) | A kind of method and device for motion picture soundtrack | |
| CN107993636B (en) | Recursive neural network-based music score modeling and generating method | |
| Nakamura et al. | Statistical piano reduction controlling performance difficulty | |
| CN111613199B (en) | MIDI sequence generating device based on music theory and statistical rule | |
| CN109448795A (en) | The recognition methods of circRNA a kind of and device | |
| CN111583891A (en) | Automatic musical note vector composing system and method based on context information | |
| CN109326270A (en) | Generation method, terminal device and the medium of audio file | |
| CN106205572B (en) | Sequence of notes generation method and device | |
| Siphocly et al. | Top 10 artificial intelligence algorithms in computer music composition | |
| Chen et al. | Robotic musicianship based on least squares and sequence generative adversarial networks | |
| Wang et al. | An intelligent music generation based on Variational Autoencoder | |
| CN110232118A (en) | A kind of novel answer preference pattern based on GRU attention mechanism | |
| CN109471951A (en) | Lyrics generation method, device, equipment and storage medium neural network based | |
| CN108074554A (en) | The evaluation method and system of a kind of piano performance | |
| CN108460111A (en) | Personal character modeling based on dialogue and generation method and device | |
| CN113488005A (en) | Musical instrument ensemble method and computer-readable storage medium | |
| Ajoodha et al. | Using statistical models and evolutionary algorithms in algorithmic music composition | |
| CN115206270A (en) | Training method and training device of music generation model based on cyclic feature extraction | |
| CN110222190A (en) | Data enhancement methods, system, equipment and computer readable storage medium | |
| Mugica et al. | A Fuzzy Inductive approach for rule-based modelling of high level structures in algorithmic composition systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20190920 |