EUROSPEECH 2003 - INTERSPEECH 2003
In this paper we present a simple algorithm for speech syllabification. It is based on the detection of the most relevant energy maximums, using two different energy calculations: the former from the original signal, the latter from a low-pass filtered version. The system requires setting appropriate values for a number of parameter. The procedure to assign a proper value to each one is reduced to the minimization of a n-variable function, for which we use either a genetic algorithm and simulated annealing. Different estimation of parameters for both Italian and English was carried out. We found the English setting was also suitable for Italian but not the reverse.
Bibliographic reference. Petrillo, Massimo / Cutugno, Francesco (2003): "A syllable segmentation algorithm for English and italian", In EUROSPEECH-2003, 2913-2916.