Automatic Analysis And Application On Beat Tracking Of Audio

With the rapid growth of Internet, Music play more important role in our daily life. Beat is the basement of music structure. It is widely use of the beat information in many application, such as music analysis and visualization, and content based music information retrieval.This paper investigates a pragmatic approach to onset detection and beat tracking. Then we investigate the possibility of using the method of beat tracking to voice data. The main work is followed:Firstly, we propose the method of onset detection with the features of music signal, such as MFCC (Mel-Frequency Cepstrum Coefficients), LPCC (Linear Prediction Cepstral Coefficients) to substitute acoustic energy. The method is more fit for the characteristic of acoustic frequency. Then we extend the work by introducing Chroma the ordinary feature of music to onset detection.Then we propose the method of music beat genre classification with MFCC, LPCC feature, and using the autocorrelation method to calculate the period of one beat. Based on these, a new means to beat tracking is presented which use the technology of Artificial Neural Network and PSOLA.Finally, we construct the system of beat tracking of music, and the effect of different acoustic feature on music beat tracking is compared. The investigation testify the fitness and accuracy of the method we proposed. And we investigate the possibility of application on voice signal. The result of experiment proved that the method of beat tracking could be used to eliminate voice beat efficaciously.