一种有效的音频信息检索技术

A New Effective Method on Audio Information Retrieval

音频数据检索是数字化信息检索的迫切需要,但国际上对音频检索技术的研究尚远未尽人意。提出了一种新的音频检索机制,利用小波变换来产生音频数据的特征向量,通过关联规则挖掘发现音频数据特征元素向量与音频数据所属类别之间的联系,从而进行音频的分类和检索。实验表明,这种方法具有较高的检索效率,能大大缩短计算时间,具有良好的检索性能。

With the growth of Internet and storage capability in recent years,multimedia data retrieval is getting more and more important.It is crucial to develop audio data retrieval technique for searching interested audio data.In this paper we propose a method that utilize the wavelet subband ,a certainly proportional high-frequency coefficient of the wavelet subband and all the approximately coefficient to form the feature vector of the audio data.By using feature vector,the proposed approach can find associations between feature vector and class attribute of the audio data and use the association rules to make more further search and retrieval.The experiment shows that this method has a high retrieval precision and search efficiency.