Musical meter estimation and music transcription

This paper concerns the segmentation and automatic transcription of acoustic musical signals. First, a method is proposed for estimating the temporal structure of music, the musical meter. The method consists of an acoustic analysis front-end and of a subsequent probabilistic model which performs joint estimation of the meter at three time scales: beat (tactus), measure, and tatum pulse levels. Secondly, musicological models are described which utilize musical information to improve transcription accuracy. 1