CN106652984B

CN106652984B - Method for automatically composing songs by using computer

Info

Publication number: CN106652984B
Application number: CN201610884621.XA
Authority: CN
Inventors: 张文铂
Original assignee: Individual
Current assignee: Individual
Priority date: 2016-10-11
Filing date: 2016-10-11
Publication date: 2020-06-02
Anticipated expiration: 2036-10-11
Also published as: CN106652984A

Abstract

The invention discloses a method for automatically composing songs by using a computer, which particularly comprises three steps of computer composition, computer composition and computer singing, and is realized by computer composition software: firstly, inputting any line of lyrics by a user, separating each line of lyrics by a separator, and operating a composition module by an algorithm to generate a melody MIDI score file; the music editing module is operated to add accompaniment tracks to the corresponding MIDI music score file and render the accompaniment file into a high-quality WAVE audio accompaniment file by using a sound source library; then, synthesizing the MIDI music score file and the appointed voice source file through a music synthesis algorithm to generate a main melody; and finally, synthesizing the main melody and the accompaniment file into the final song. The invention is a great innovation in the field of computer music, and ordinary people can convert own characters into songs which have melodies, have customized accompaniment and can be sung within a few seconds.

Description

Method for automatically composing songs by using computer

Technical Field

The invention belongs to the technical field of song creation methods, and particularly relates to a method for automatically creating songs by using a computer.

Background

The automatic song creation is divided into three techniques in the flow, namely computer composition, computer editing and computer singing. Computer composition, or algorithmic composition, is a study that attempts to use some formalized process to minimize the level of intervention by an average user or composer in the composition of music using a computer. The main techniques adopted by the traditional automatic composition algorithm in the prior art are as follows: markov chains, stochastic processes, hierarchical techniques, knowledge base systems, musical grammars, artificial neural network techniques, and genetic algorithms, all simulate musical melodies using machine learning algorithms. Because music is an artwork which is difficult to quantify, it is inconvenient to adopt a mechanical learning algorithm for simulation, i.e. the training result of the model cannot be accurately evaluated, because the good hearing degree of music is very difficult to quantify, the music is a unique perceptual knowledge of human beings, and the evaluation of different people also has deviation. Therefore, music made by the traditional machine learning method is difficult to conform to the aesthetic of people, thereby losing the artistry.

The computer music composing technology belongs to the subject with very strong music speciality, and because the requirement of algorithm development on music knowledge is very high, no mature technical implementation scheme exists in the field of computers at present. Meanwhile, the algorithm technology has the same strict requirement on computer knowledge, and no automatic music composition solution exists in the field of music creation, so that the computer music composition method is a new attempt and innovation in the field. The computer singing technology is a sound synthesis technology, specifically, according to a specified music score, a preset sound source file is used, and a sound with a specified pitch and duration is constructed through audio frequency modulation and amplitude modulation operation.

Disclosure of Invention

The invention provides a method for automatically composing songs by using a computer, which can convert characters into music in several seconds in order to overcome the defects of the prior art.

The invention is realized by the following technical scheme: a method for using the computer to compose the song automatically, include three steps of computer composition, computer song editing and computer singing specifically, the invention is realized through the computer song editing software; the computer composition mainly takes a chord trend theory in a music theory as a basic frame and combines a machine learning algorithm to calculate parameters; the computer music editing is that the music track addition is carried out according to the MIDI file of the main melody and the appointed chord progression mode and the musical instrument with common timbre; the computer singing, namely music synthesis, is the step of synthesizing the main melody and the accompaniment into the final song;

the specific process of automatically composing songs by using a computer is as follows: 1) computer composition: firstly, inputting any line of lyrics by a user, separating each line of lyrics by a separator, and operating a composition module by an algorithm to generate a melody MIDI score file; 2) and (3) computer editing: the music editing module is operated to add accompaniment tracks to the corresponding MIDI music score file and render the accompaniment file into a high-quality WAVE audio accompaniment file by using a sound source library; then, synthesizing the MIDI music score file and the appointed voice source file through a music synthesis algorithm to generate a main melody; 3) singing by a computer: and finally, synthesizing the main melody and the WAVE audio accompaniment file to finally synthesize a complete song WAVE audio file.

In the computer composition step: the algorithm data preparation needs to be done first (no repetition is needed in the subsequent runs): collecting several popular songs in various tones, analyzing chord progression mode (for example 1645) of each song, finding chord tone combination (such as 135 in C major tone) in melody, and scoring each song with sound level in manual mode; using a supervised learning algorithm, taking the chord tone combination condition of each song as input, taking the listening degree of each song as a supervision signal, and finally calculating various optimal chord tone combination sequences in each mode; the steps are used for preparing early-stage data of the algorithm, and the composition grace degree of the algorithm is continuously improved along with the increase of the number of training songs.

The specific calculation logic of the computer composition algorithm is as follows: the method comprises the steps that a user inputs lyrics, the number of the lyrics and the number of words of each sentence are analyzed through an algorithm, the lyrics and the number of words of each sentence are used as input and are transmitted into a composition algorithm, the composition algorithm randomly performs optimal chord tone combination matching according to different word numbers of each sentence, random notes close to the chord tones are added around each chord tone, unique melodies are generated for each lyric, the melodies of all the phrases are combined together and are expressed through a MIDI format, and then the MIDI file of the main melody of the composition is generated.

The computer music composition adopts different adding modes according to the characteristics of different timbres of four major pianos, guitars, bass and drums, such as guitars and bass, and needs to carry out chord configuration according to the chord mode, and the drums need to carry out automatic chord configuration according to the rhythm of MIDI files. Taking guitar track addition as an example for explanation, the specific operation method is as follows: the method comprises the steps of obtaining rhythm and pitch information of a main melody by analyzing a MIDI file into a score format, analyzing chord trend corresponding to each measure of the main melody, and adding harmony syllables (such as domiso) at positions corresponding to each beat of each measure by taking a piano, a guitar, a bass or a drum as timbre. The accompaniment of a one-rail piano, guitar, bass or drum can be obtained after the addition of all the materials is finished. According to different style selections, different orchestrator combinations are matched with songs in different styles through an algorithm, for example, popular styles are added, music tracks such as a grand piano, an electric guitar, a bass and a drum are added, notes are added to each music track according to the method, complete accompaniment note information is obtained after the completion, the note information is exported in a MIDI format, and finally, the note information is rendered into a high-quality audio file through a soft sound source, namely, the high-quality audio file is stored as a WAVE format file.

The specific process of computer singing, namely music synthesis, comprises the following steps: firstly, obtaining an output MIDI file of a music composing module, and analyzing the file to obtain appointed music score note information; then the program reads the pre-recorded voice source file, according to the music score, uses SMS (short message service) algorithm to construct the audio frequency, constructs the music according to the pitch and time appointed by the music score; and finally, mixing the music serving as the main melody of the song and the accompaniment WAVE file to obtain a final multi-track song file.

The invention has the beneficial effects that: the invention combines the principles of the related documents in the prior art to develop a melody and accompaniment track interface in the form of MIDI of an electronic score, which can create music more convenient for synthesizing the appointed melody and accompaniment. The method for automatically creating songs by using a computer is a great innovation in the field of computer music, marks that the common public can also participate in music creation due to the song creating method, and can convert own characters into the songs which have melody, have customized accompaniment and can be sung within a few seconds by a common person through automatically creating the songs by using the computer, thereby really realizing the purpose of converting the characters into the music within a few seconds.

The invention can adjust various parameters according to the self feeling to create different styles of music, the randomness of the algorithm of the invention ensures that two pieces of music which are completely the same cannot be created basically, and the generation process of each song is unique. Because the number of people is ten million times of professional musicians, the method can create an unthinkable number of songs, and even the art is a large random rule discovery process, so that the method for creating the songs is very likely to discover a new music artistic form from a large number of music works.

Drawings

Fig. 1 is a functional block diagram of software programming in the method of automatically composing songs using a computer according to the present invention.

Detailed Description

The invention is further described below with reference to the accompanying drawings and specific examples.

A method for using computer to automatically compose songs specifically comprises three steps of computer composition, computer composition and computer singing, which are realized by computer composition software; the computer music composition mainly takes a chord trend theory in a music theory as a basic frame and combines a machine learning algorithm to calculate parameters; the computer music editing is that the music track addition is carried out according to the MIDI file of the main melody and the appointed chord progression mode and the musical instrument with common timbre; the computer singing, namely music synthesis, is the step of synthesizing the main melody and the accompaniment into the final song;

The specific process of computer singing, namely music synthesis, comprises the following steps: firstly, obtaining an output MIDI file of a music composing module, and analyzing the file to obtain appointed music score note information; then the program reads the pre-recorded voice source file, according to the music score, uses SMS (Spectral modeling Synthesis) and other algorithms to construct the audio frequency, constructs the music according to the pitch and time appointed by the music score; and finally, mixing the music serving as the main melody of the song and the accompaniment WAVE file to obtain a final multi-track song file.

Example (b): as shown in fig. 1, the detailed creation process of a song in the method for automatically creating songs using a computer according to the present invention is as follows: 1) the composition engine module: simulating composition by using an artificial neural network algorithm and a related music theory, and calculating a random rhythm and melody pitch combination according to a specified chord trend rule so as to generate a melody with a specified bar number; 2) a composition engine module: randomly generating accompaniment tracks with different timbres in a rule according to the chord trend of the melody, such as a piano, a guitar, a bass, a drum and the like, and performing flower adding processing according to different musical instrument characteristics; 3) the singing engine module: and synthesizing sound of the sound source file according to the specified music score by a speech synthesis technology in natural language processing, and outputting a waveform file.

Finally, it should be noted that the above-mentioned contents are only used for illustrating the technical solutions of the present invention, and not for limiting the protection scope of the present invention, and that the simple modifications or equivalent substitutions of the technical solutions of the present invention by those of ordinary skill in the art can be made without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for automatically composing a song using a computer, characterized by: the method for automatically composing songs by using the computer specifically comprises three steps of computer music composition, computer music composition and computer singing, and is realized by computer music composition software; the computer composition mainly takes a chord trend theory in a music theory as a basic frame and combines a machine learning algorithm to calculate parameters; the computer music editing is to add the music track according to the MIDI file of the main melody and the appointed chord and the musical instrument with the common tone; the computer singing, namely music synthesis, is a step of synthesizing the main melody and the accompaniment into a final song;

the specific process of automatically composing songs by using a computer is as follows: 1) computer composition: firstly, inputting any line of lyrics by a user, separating each line of lyrics by a separator, and operating a composition module by an algorithm to generate a melody MIDI score file; 2) and (3) computer editing: the music editing module is operated to add accompaniment tracks to the corresponding MIDI music score file, and the accompaniment file is rendered into a high-quality WAVE audio accompaniment file by using a sound source library; then, synthesizing the MIDI music score file and the appointed voice source file through a music synthesis algorithm to generate a main melody; 3) singing by a computer: finally, the main melody and the WAVE audio accompaniment file are synthesized, and finally, a complete song WAVE audio file is synthesized;

in the computer composition step of step 1): firstly, algorithm data preparation is needed, a plurality of popular songs of various tones are collected, the chord progression mode of each song is analyzed, chord and tone combination in the melody is found out, and each song is scored according to the hearing degree in a manual mode; using the machine learning algorithm, taking the chord tone combination condition of each song as input, taking the listening degree of each song as a supervision signal, finally calculating various optimal chord tone combination sequences in each mode, and continuously improving the composition grace degree of the algorithm along with the increase of the number of training songs;

the specific calculation logic of the computer composition in the step 1) is as follows: the method comprises the steps that a user inputs lyrics, the number of the lyrics and the number of words of each sentence are analyzed through an algorithm, the lyrics and the number of words of each sentence are used as input and are transmitted into a composition algorithm, the composition algorithm randomly performs optimal chord tone combination matching according to different word numbers of each sentence, random notes close to the chord tones are added around each chord tone, unique melodies are generated for each lyric, the melodies of all the phrases are combined together and are expressed through a MIDI format, and then the MIDI file of the main melody of the composition is generated.

2. A method of automatically composing a song using a computer according to claim 1, wherein: in the computer-based composing step of step 2): according to the characteristics of different timbres of the four major pianos, guitars, bass and drums with the music, different adding modes are adopted; analyzing a MIDI file into a music score format to obtain rhythm and pitch information of a main melody, analyzing chord trend corresponding to each measure of the main melody, adding harmony syllables at positions corresponding to each beat of each measure by taking a piano, a guitar, a bass or a drum as timbre, and obtaining a track musical instrument accompaniment after all additions are finished; according to different style selections and algorithms, different orchestrator combinations are matched for songs in different styles, notes are added to each track according to the method, complete accompaniment note information is obtained after completion, the note information is exported in a MIDI format, and finally the note information is rendered into a high-quality audio file through a soft sound source, namely the high-quality audio file is stored as a WAVE format file.

3. A method of automatically composing a song using a computer according to claim 1, wherein: the specific process of singing by the computer in the step 3) is as follows: firstly, obtaining an output MIDI file of a music composing module, and analyzing the file to obtain appointed music score note information; then program reads the pre-recorded voice source file, according to the music score, uses computer to make audio frequency structure, constructs the music according to the pitch and time appointed by the music score; and finally, mixing the music serving as the main melody of the song and the accompaniment WAVE file to obtain a final multi-track song file.