Acoustic features for detection of aspirated stops

Abstract

Aspiration is an important phonemic feature in several Indian languages. Unlike English, languages such as Marathi have lexicons in which words with different meanings differ only in the aspiration feature of the initial voiced or unvoiced stop. Thus the reliable discrimination of aspirated stops from their unaspirated counterparts is important in automatic speech recognition for such languages. The important acoustic distinctions include durational features as well as fine spectral structure features. Traditional frame-based spectral representations such as MFCCs used in HMM-based recognizers do not explicitly encode these cues. In this work, we explore various acoustic features for aspiration detection in voiced and unvoiced stops of Marathi. Enhancements to available methods of aspiration detection borrowed from voice quality measures are found to provide improved detection of phonemic aspiration in stops. The performance of a landmark-based acoustic feature classifier is compared with MFCC-HMM baseline system for the recognition of aspirated and unaspirated stops.