An audio fingerprinting extraction algorithm based on lifting wavelet packet and improved optimal-basis selection

Audio fingerprinting technology is widely applied to the analysis and processing of digital signal, especially in the application of speech recognition which is one of the most popular fields of the intelligent multimedia and artificial intelligence. Traditional audio fingerprinting extraction algorithm is based on the decomposition and reconstruction of the wavelet packet. But the requirement of computational capacity and memory is so large. So this paper proposed an algorithm which is based on the lifting wavelet packet and the improved optimal-basis selection to find the coefficient of optimal wavelet packet. Then the average of the logarithmic energy entropy is adopted as the characteristic parameter. And the capacity of computing and memory is better than the traditional algorithm because of the lifting wavelet packet which is more suitable for processing of speech online and the design of intelligent multimedia. And the experiment results indicate that this algorithm is not only robust for the audio which is handled by some kinds of methods and can reflect the overall characteristics of the audio very well, but also has good distinguishability between different audio.