Solving The Satisfiability Problems With Evolutionary Negative Selection Algorithms

The semantic information is the top abstract form of musical content. It can be understood and exchanged by people directly. The emotional semantic meaning is the essential characteristic for music expression. The research for musical content analysis and composition based on emotion is the critical component of computer music. It will promote the emotional effect on technical of multimedia and music entertainment.The content of this paper focuses on some key problems of musical content analysis and composition. It deeply investigates how to integrate the utilization of measurement and features for similarity estimation, how to make up the difference between similarity of musical bottom features and similarity of high level semantic information of emotional apperceive, and how to solve the problems of expression of musical knowledge, creativity, human-computer interaction and quality evaluation of the generated music?The contribution of this paper can be concluded as followings:(1) The midi format for music is adopted. The processes of locating the main track, distinguishing the main frame have been done. A subjective experiment has been designed to estimate the emotion of the main frame. The aim is to find the main frame with independent emotion, and this independent emotion is used as category of the main frame. The statistical features and sequential features of the note attributes on the main frame have been extracted to construct the feature space of music. Those two results are used to build the data set with uniform format for the following experiments.(2) In the process of constructing the model of musical content analysis, a new method, measurement and feature selection using Genetic Algorithm, has been proposed to make an integral utilization of measurement and features for pattern similarity estimation. The purpose is to search for the optimal combination of measure function and feature weights. The experiments on artificial data set and partial UCI data sets have been put into practice to verify the effectiveness of the proposed method. After that, another method, Music Perceptual Similarity Estimation Using Interactive Genetic Algorithm, has been proposed based on the above method for considering subjectiveness of the musical emotion. The human's subjective perceptual similarity estimation is used as the fitness function in genetic algorithm to make up the difference between similarity estimation of music bottom level features and similarity estimation of high level semantic information of emotional apperceive. Finally, the model, whichi is satistied with human's subjective perceptual similarity estimation is built with the musical data set above.(3) The KTH rules is used to represent the musical knowledge and the Interactive Genetic Algorithm is used for the emotional music generation. In the process of music generation, the effectiveness of the generated music is estimated by the user with the difference between subjective perception and emotional requirement predefined. This measrue merge user's subjective perception into the generated music. The mothod from psychophysics of constructing psychological scale is used to verify the quality of the generated music. It shows a new approach for quality estimation for generated music.