Improved onset detection algorithm based on fractional power envelope match filter

In music and speech recognition, onset detection plays an important role for extracting the note of a music signal or the syllable of a speech signal. There are several existing onset detection algorithms, such as the difference of magnitude method, the short-term energy method, the surf method, and the high frequency content method. In this paper, we proposed an improved onset detection algorithm, which mainly applies the techniques of the fractional power amplitude, the envelop-matched filter, and other techniques to improve the accuracy and the efficiency of onset detection for humming signals. From simulations, our proposed method has both less computation time and higher accuracy than the existing methods. It also obviously increases the hit rate of the query-by-humming system.