摘要： 数字音乐指纹提取的主要目的是建立一种有效机制，用于比较2个音乐文件的听觉质量。提出一种基于小波包最优基分解的音乐指纹提取算法，利用与音频内容密切相关的小波包系数，将其作为特征进行指纹提取。实验结果表明，该算法对MP3, WMA和RM压缩、噪声、Stirmark for audio工具中常见的音频信号处理具有强鲁棒性，且在不同音乐之间具有较高可区分性。

关键词: 数字音乐指纹, 小波包变换, 最优基, 鲁棒性

Abstract: The main purpose of digital music fingerprint extraction is to establish an effective mechanism used to compare the auditory quality between two pieces of audio. This paper presents a music fingerprint extraction algorithm based on wavelet packet best-basis decomposition. It uses wavelet packet coefficients related to radio content as the feature to extract fingerprint. Experimental results show that this algorithm is robust against common audio signal operations like MP3, WMA and RM compression, noise addition, and audio processing in Stirmark for audio. This algorithm exhibits high ability to differentiate between different songs.

Key words: digital music fingerprint, Wavelet Packet Transform(WPT), best-basis, robust

中图分类号: 

• TN912.3