ISCA Archive - Speech fundamental period estimation by a novel parallel processing method (original) (raw)

Speech fundamental period estimation by a novel parallel processing method

Ian S. Howard

The algorithm described here is a time domain speech fundamental period estimator. Its operation involves filtering the speech with a filterbank. The filterbank outputs are then represented in terms of positive-going zero-crossing locations and their envelopes. Finally the period epoch marker locations are "optimally" selected. The algorithm is compared with a reference based on the output of a laryngograph.