Musical Audio Beat Tracking using Hidden Markov Model

The proposed method describes the beat tracking system of musical audio signals. Here, the beat tracking deals with the explicit modeling of non-beat states and estimates the time between consecutive beat events. The expected accuracy of the estimated beats is also provided additionally. To predict the accuracy of the beat estimate a k-nearest neighbor regression algorithm is used. A database of 222 musical signals of various genres is used to statistically evaluate the performance for beat tracking system. A new perspective for beat detection is presented which can be used to enhance the performance of the proposed algorithm and relates it with the human tapping.