Research On Intelligent Music Composition Based On Rules And Probabilistic Models

With the advance of science and technology,artificial intelligence is also developing rapidly.Great progress has been made in the combination of artificial intelligence and different fields.In the aspect of intelligent composing,the main methods of combining artificial intelligence include: machine learning method based on artificial neural network and related Markov chain,genetic algorithm and hierarchical structure etc.The main problem of current intelligent composing is that the method is too precise and technical,which is not entirely consistent with the fuzzy thinking of music.The output is either high predictable or random,resulting in less beautiful or euphonic music.To make the output of intelligent composing not too messy,nor too monotonous,and make it more euphonic,this paper proposes an intelligent composing method fusing rules and probabilistic models.Firstly,the MIDI features and related data are extracted from music clips in the corpus with the assistance of corresponding musical knowledge to solve the information expression problem of music scores.After that,the constraint satisfaction problem is analyzed and the pitch generation rules of different degrees are set.Based on the rules,the existing probabilistic models are improved to ensure that the output pitch sequence conforms to the constraints.Then the duration analysis is completed and a probability model of rhythm is constructed.Based on that model,probability analysis is carried out and the duration sequence which conforms to the constraint satisfaction is generated to form the rhythm.Finally,a new music melody is generated by the fusion of pitch sequence and duration sequence.An artificial auditioning experiment on a certain number of output compositions shows that such method can improve the aesthetic feelings of the output results.To a certain extent,this study can solve the problem that the output melody is too random or monotonous.