In popular music, the classification of the traditional concept of style, must rely on artificial given style labels, this classification of qualitative level exist great controversy, and low efficiency. Therefore, in this paper, considering the establishment of mathematical model, using the pattern recognition algorithm is appropriate, with the help of computer automatic classification of popular music.
In this paper, the first is to extract feature vectors of pop music. Through access to information that, popular music between the difference lies mainly in the climax, flat part are similar, but the climax of data frame energy, flat part of the data frame energy. Therefore, using the data frame energy as the characteristic, also can use the frame energy size filtering of the flat part of the data; second, data frame energy value just reflect the characteristics of the numerical size, does not reflect the distribution of energy, so consider the data frame energy ratio as second features. Therefore, the feature vector of popular music from the frame energy and the frame energy ratio. In order to verify the correctness of the extracted feature vectors, using the current popular songs are validated, the results show, the extracted feature vectors can be a very good identification of pop music, proves the rationality of the extracted feature vector.
The second work is based on the feature extraction, classification of the pop music using cluster analysis method. First, using the clustering algorithm based on shortest distance, cluster analysis procedure is as follows:
Every pop song alone as a class;
According to the principle of minimum distance, in turn elect two songs, and a new class of;
If a piece of music has been attributed to a class, put another song into the class; if the two music just has to belong to class two, class two and put it into a class; each time the merge, are crossed out and sequence of the music in the same row;
Then, after you can put all the music is classified as a class, so you can according to merge to the sequence clustering dendrogram, to complete the classification.
Use of the algorithm of pop music classification results, the accuracy rate of algorithm is high, but the algorithm used in the simple data, losing some information, there can be room for improvement. Therefore, fuzzy C means clustering method is then used in this article, the shortest distance clustering algorithm is improved, the calculating example proved that, the improved classification accuracy rate of pop music, can reach 84.67%.
Finally, according to the calculation results, advantages and disadvantages of the algorithm are summarized, and the algorithm of the corresponding promotion.