用小波包变换产生音频数据索引的方法

针对音频多媒体数据库中基于例子和基于内容的查询,本文提出一种产生音频数据索引的方法.这里,我们首先讨论了小波包分解的过程和最好基及代价函数的选择方法.其次,对现有的用小波变换产生音频数据索引的二个方法进行比较,并提出基于小波包最好基变换产生音频数据索引的方法.再次,我们提出用音频数据的小波包最好基变换系数的部分最高值的能量作为音频数据索引.最后,我们把这种方法与直接采用小波变换产生索引的方法相比较.实验结果表明这种新方法具有较高和较稳定的检索精度.

Method of Using the Wavelet Packet Transform Derivation Audio Data Index

A method for deriving the audio data index is proposed in this paper,which aim at query-by-example and query-by-content in audio multimedia databases.Here,first we discuss the process of wavelet packet decomposition and the method for choosing the best base and cost function.Second we prepare two schemes,which has been used,that deal with deriving the index of wavelet transform.And we propose the method to derive the audio data index that it is based on the best base transform in the wavelet packet.In the third,we take the part of energy for the audio data index which come from the highest value coefficients of the best base transform in the wavelet packet.In the last,we take our method comparing with the method that derived the index directly form the wavelet transform.The experiment shows that there are higher and robust for retrieving precision in the new method.