音频指纹搜索中数据预处理的改进算法

Improved algorithm of data preprocessing in audio fingerprint search

不同的音频指纹提取算法需要不同的音频指纹搜索。针对某些特定(例如:foosic算法)的音频指纹提取算法,在数据预处理方面提出并实现了两种新的音频指纹搜索算法:PCA(Principle Component Analysis)主成分分析算法、不同帧之间相应的数据求和算法。实验结果表明:应用PCA算法,在搜索正确率为94.98%的情况下,搜索时间缩短为8.42%;应用求和算法,在搜索正确率为95.92%的情况下,搜索时间缩短为3.72%。

Different Audio Fingerprint（AF） feature extraction algorithms require different AF searching algorithms.For some specific AF feature extraction algorithms（e.g.foosic algorithm）,in data processing,two new AF searching algorithms are proposed and implemented：Principle Component Analysis（PCA） algorithm and sum of the corresponding data between different frames algorithm.The experimental results show that applying PCA algorithm,the accuracy is 94.98% while the search time is as low as 8.42%;applying sum algorithm,the accuracy is 95.92% while the search time is as low as 3.72%.