Abstract

This paper concerns both rhythm recognition and tempo analysis of expressive music performance based on a probabilistic approach. In rhythm recognition, the modern continuous speech recognition technique is applied to find the most likely intended note sequence from the given sequence of fluctuating note durations in the performance. Combining stochastic models of note durations deviating from the nominal lengths and a probabilistic grammar representing possible sequences of notes, the problem is formulated as a maximum a posteriori estimation that can be implemented using efficient search based on the Viterbi algorithm. With this, significant improvements compared with conventional “quantization ” techniques were found. Tempo analysis is performed by fitting the observed tempo with parametric tempo curves in order to extract tempo dynamics and characteristics of performance to use. Tempo-change timings and parameter values in tempo curve models are estimated through the segmental k-means algorithm. Experimental results of rhythm recognition and tempo analysis applied to classical and popular music performances are also demonstrated.

Citations