Support Vector Machine Integration And Application In The Music Category

With the rapid development of communication, computer and Internet, various information increases exponentially. There are more great opportunities for people to have access to the large quantities of multimedia contents, such as images, videos and audio. These multimedia contents have gradually become the main form of information media in the field of information process. But since the fast growing of the data volume, how to manage the contents automatically has emerged as an urgent problem. Especially to the all kinds of music signals around us, fast and efficient methods are required to classify and manage them (according to different styles or singers).Music classification has always been much accounted of by people as one of the voice recognition problem. Along with the fast developments of voice recognition technology, many algorithms and methods have been applied in this area.However,it's still far from the large-scale application of music auto classification as the variety and complicacy of music. Most of the contemporary algorithms for audio signal classification include two stages: feature extraction stage and classification stage. Lots of music features can be applied to implement this algorithm, including the short-time energy and short-time zero-crossing-rate etc.from the time domain, the bandwidth and brightness etc.from the frequency domain, also the MFCC (Mel-frequency cepstral coefficients) coefficient which is based on the perception. And the many high efficient algorithms in the Pattern Recognition and Pattern Classification such as Gaussian Mixture Model(GMM)[37],Neural Network(NN),Hidden Markov Mode(HMM)[10][12][21][23][24], etc. can be utilized to implement the classification. When facing such many features and classification algorithms, how to combine them to achieve a better classification accuracy rate? Is it possible to do some preprocessing on some of the features or do some optimization on the classifiers base upon the specialty of music classification to achieve a higher classification accuracy rate? To answer these questions, this thesis proposes a new music classification method base on the theory of Ensemble and the Support Vector Machines.Support Vector Machines (SVM) has been applied in many fields and achieved plentiful fruits already since proposed by Vapnik in 1995.Based on statistical learning theory (SLT),SVM possesses many merits such as concise mathematical form, standard fast training algorithm and excellent generalization performance, so it has been widely applied in data mining problems such as pattern recognition, function estimation, time series prediction and classification, ect. However, some problems, for example, the model selection, efficiency of SVM for large-scale training set, etc, still need to be solved in SVM research. Generally, almost all researches use single SVM as learner, and multi-SVM learner methods are scarce thought out. Ensemble learning technology as an effective multi-learner method has been obtained many valuable achievements. If the ensemble learning technology can be introduced to SVM, the generalization performance of SVM may be improved efficiently. Therefore, research on ensemble SVM learning becomes an important research issue.In the thesis, ensemble SVM learning method and the principle and method of music classification are investigated systematically. And the now existing algorithm of ensemble SVM is improved and applied in music classification methods. In the last place are the experiments for the algorithm performance evaluation. In the experiments, the simulations are performed on different feature sets using different classifiers, and the results not only verify the truth that the classification accuracy rate improves a lot after using the ensemble SVM, also show clearly the advantage of ensemble SVMs over those traditional classifiers in the field of music classification.