Improved DYPSA Algorithm for Noise and Unvoiced Speech

Abstract

The DYPSA algorithm detects glottal closure instants (GCI) in speech signals. We present an improvement in the algorithm in which a voiced/unvoiced/silence discrimination measure is applied in order to reduce the spurious GCIs detected incorrectly for noise and unvoiced speech. Speech classification is addressed by formulating a decision rule for the glottal closure instant candidates which classifies the candidates as voiced or non-voiced on the basis of feature measurements extracted from the speech signal alone. The technique of Dynamic Programming is then employed in order to select an optimum set of epochs from the GCI candidates. The algorithm has been tested on the APLAWD speech database with 87.23% improvement achieved in reduction of spurious GCIs.