基于关键段和小波包最优基的流行歌曲索引

摘要

为了对大量流行歌曲数据进行有效索引,提出了基于流行歌曲关键段和小波包最优基的索引方法.首先对流行歌曲关键段提取小波包最优基,其次分别计算最优基中各重要结点所有系数的能量作为关键段的特征向量,然后根据重要结点个数找到相应的子数据库,最后采用均方误差指标在子数据库中进行数据匹配,实现流行歌曲的索引.实验中对不同流行歌曲关键段进行了索引实验,验证了该方法的可行性.实验结果表明该方法充分考虑到流行歌曲的特点,搜索精度比较高.%This paper presents an index method for the popular songs based on the key segment and wavelet packet best-basis. The wavelet packet best-basis is extracted from the key segment of popular songs. Based on each important point the energy of all coefficients in the best-basis is calculated as the feature vector of the key segments. The corresponding sub-database is found according to the number of important points. MSE(mean squared error) target is used to carry out the data match in the sub-database. The key segments of different popular songs are used in the index experiment. Experimental results show that the searching precision of this method is quite high.