Fundamentals of speech recognition
Fundamentals of speech recognition
Automatic Segmentation of Acoustic Musical Signals Using Hidden Markov Models
IEEE Transactions on Pattern Analysis and Machine Intelligence
Score following: state of the art and new developments
NIME '03 Proceedings of the 2003 conference on New interfaces for musical expression
Hi-index | 0.00 |
In this paper we present algorithms for the automatic time-synchronization of score-, MIDI- or PCM-data streams which represent the same polyphonic piano piece. In contrast to related approaches, we compute the actual alignment by using note parameters such as onset times and pitches. Working in a score-like domain has advantages in view of the efficiency and accuracy: due to the expressiveness of score-like parameters only a small number of such features is sufficient to solve the synchronization task. To obtain a score-like representation from the waveform-based PCM-data streams we use a preprocessing step to extract note parameters. In this we use the concept of novelty curves for onset detection and multirate filter banks in combination with note templates for pitch extraction. Also the data streams in MIDI- and score-format have to be suitably preprocessed. In particular, we suggest a data format which handles possible ambiguities such as trills or arpeggios by introducing the concept of fuzzy-notes. Further decisive ingredients of our approach are carefully designed cost functions in combination with an appropriate notion of alignment which is more flexible than the classical DTW concept. Our synchronization algorithms have been tested on a variety of classical polyphonic piano pieces recorded on MIDI- and standard acoustic pianos or taken from CD-recordings.