Objective Description And Quantitative Similarity Evaluation Of Music Based On Mathematical Model

Generally,different styles of music created by different composers have obvious differences in the characteristics of the symbolic(non-audio)music information(such as pitch,rhythm,etc.).At present,there is room for improvement of the effect of similarity measurement on chord collocation melody in the task of distinguishing different styles of symbolic music in the research field of objective quantification evaluation of similarity of human composition or artificial intelligence(AI)composition.At the same time,there is currently no method for quantitatively evaluating the difference of the dependency of the melody pitch and the melody rhythm in different styles of music created by different composers(including AI),and a method to quantitatively evaluate the multi-layered dependence of chord-melody pitch-melody rhythm of different styles of music created by different composers(including AI).In addition,at present,there is still a lack of quantitative research on the similarity between different styles of human composition and corresponding AI composition.In order to solve the above problems,based on the Hidden Markov Model and the Hierarchical Hidden Markov Model,this paper proposes a variety of objective music description methods and quantitative evaluation methods of the similarity of music,and quantitatively evaluates the differences in various dependencies between chords,melody pitch,and melody rhythm between different styles of human composition and corresponding AI compositions,and between different styles of music created by different humans;this paper proposes a variety of quantitative evaluation methods of music similarity combining the dynamic characteristics and static characteristics of the time series of music,and conducts similarity quantification evaluation experiments and distinguish experiments of different styles of music from multiple views.This paper verified the following conclusions through experiments:1)After subdividing the music styles of human compositions,there is a clear and quantifiable difference in the dependence of the chord and melody pitch between the human compositions of specific different styles and the corresponding AI compositions.The difference in chord and melody pitch dependent manner in same style of music is significantly smaller than the difference between different styles of music,and the objective music description method and the corresponding quantitative evaluation method of music similarity proposed in this research can show this difference significantly.In addition,after combining single-step timing chord progression(temporal dynamic characteristics)and the dependence of chords and melody pitch(static characteristics),this research proposes a quantitative evaluation method for the similarity of the overall progress of musical harmony and this method is better for distinguishing specific different styles of music than the similar metric evaluation method of the chord melody collocation method that only involves static features in previous related research.2)In specific different styles of music,there is a clear and quantifiable difference in the dependence between the pitch of the melody and the rhythm of the melody,and the difference between the pitch of the melody and the rhythm of the melody between the music of the same style is smaller than the difference between the music of different styles.The objective music description method and corresponding music similarity metric evaluation method proposed in this research can show this difference significantly.In addition,this research proposes a method for evaluating the overall similarity of the melody based on the single-step progress of the melody pitch and the dependence between the melody pitch and the melody rhythm.This method can present the difference of the overall progress of the melody of specific different styles of music with good effect(the difference of the overall progress of the melody of the same style music is significantly smaller than the corresponding difference between the different styles of music).At the same time,this method can distinguish different styles of music with a high accuracy.3)In specific different styles of music,there are obvious quantifiable differences in the multi-layer dependence of chord-melody pitch-melody rhythm;in the music of the same style,the difference of the multi-layer dependence of chord-melody pitch-melody rhythm is obviously smaller than the corresponding difference between different styles of music.The objective music description method and corresponding music similarity metric evaluation method proposed in this research can show this difference significantly.In addition,this research proposes a method for evaluating the overall similarity of music combining the progress of single-step timing of harmony and the multi-layer dependence of chord-melody pitch-melody rhythm.This method can present the overall difference of specific different styles of music with good effect(the overall difference of the same style of music is significantly smaller than that of different styles of music).At the same time,this method can distinguish different styles of music with high accuracy.Finally,based on the research content of this paper,it is concluded that there are obvious quantifiable differences in the various dependencies between chords,melody pitch,and melody rhythm of specific different styles of music,and the differences between the same style of music are less than that of different music styles.The objective music description and similarity quantification methods proposed in this paper can significantly show the differences in different aspects of music,and can distinguish specific different styles of music created by humans or AI with a high accuracy rate.The work of this paper improves the research on the objective evaluation of the similarity of music in the symbolic domain,making it more closely related to music theory,and thus making the work results of this paper objectively more reliable.The work results of this paper can be used for the evaluation of AI composition,especially the evaluation of AI composition in different styles after the subdivision of the music style of human composition,so as to avoid factors such as music education background and personal preferences reduce the objectivity of the hearing test result in the similarity evaluation of AI composition and human composition through human hearing test.In addition,the research of this paper achieves a good music style discrimination effect under the condition of low data processing cost.Therefore,the research of this paper also have potential application value in the field of music style classification based on music big data.