A new method for extracting a period of beat of music in compressed domain of TwinVQ audio compression

A beat is primitive information of music for content-based music analysis. Scheiner defined that a beat is the sequence of equally spaced phenomenal impulses which define a tempo for the music. To extract the beat of a piece of music is to estimate the duration between impulses from a sequence of signals of the music. We call the duration the "period of beat" or "period". In this paper, we propose a new method for extracting the period of beat of the music in the compressed domain of TwinVQ audio compression (MPEG-4 audio). Our key idea is to extract the period of the beat from the set of sequences of the autocorrelation coefficients computed in the encoding step of Twin VQ audio compression or directly computed from the bit streams without decoding. For extracting the period of beat of music, we apply time-frequency analysis to the sequence of the i-th autocorrelation coefficients (i = 1, ..., 20) and determine the period of the beat. To evaluate the performance of our method, we apply the TwinVQ audio compression to the signals of 62 polyphonic music ripped from music CD tracks and get the set of sequences of the autocorrelation coefficients, and then apply our extraction algorithm to the set to extract the period of the beat. The results of experiments on 62 polyphonic music show good performance of extraction of the period of beat. Our contribution to the content-analysis of music is to realize a method for the extraction of the period of beat in the compressed domain of TwinVQ audio compression and to make a base for music structure analysis based on the beat of the music.