Audio fingerprint extraction based on time-frequency domain
Zhengzheng Liu, Cong Li, Sanxing Cao · 2016
This paper introduced the development background and analyzed the research significance of audio recognition system. Then explained the principle of digital audio fingerprinting technology and related applications, and introduced the basic concepts and properties. It also analyzed the most commonly used model and summarized the corresponding algorithm of digital audio system. Finally, several methods of extracting feature parameters are analyzed, and the method of feature extraction based on wavelet transform is proposed. In this study, it used wavelet packet technology, the coefficients of each frame of the audio file are decomposed into 3 layers, and then, used the decomposition coefficient energy, zero crossing rate, the center of mass as the audio fingerprint. In addition to the above fingerprint, the design also used the audio parameters MFCC as another audio fingerprint features.