Time variable tempo detection and beat marking

In this paper, we present a novel approach to the automatic estimation of tempo over time. This method aims at detecting tempo for music with and without percussion. An onset-energy function is first proposed based on the reassigned spectral energy flux. A combination of Discrete Fourier Transform and Frequency Mapped Autocorrelation Function is used to estimate the dominant periodicities at each time. A Viterbi algorithm is then used in order to detect the most likely tempo and meter/beat subdivision paths over time. The performance of the proposed method is evaluated on three databases.