Abstract

Speech waveform segments can roughly be categorized as voiced or unvoiced, in accordance with periodicity properties of the glottal source, and inferring this classification from data is in turn an important task underlying a variety of speech classification problems. This presentation describes a formal hypothesis testing framework for the detection of periodicity in general acoustic sources, with application to online voiced/unvoiced segmentation of speech signals. Beginning with the classical approach of Fisher, a variety of test statistics are proposed and analyzed in this context. Asymptotic analyses are provided, along with simulations to demonstrate the efficacy of such methods in the presence of compound periodicities, harmonic structure, and the low signal‐to‐noise‐ratio environments typical of real‐world speech applications. [Work supported in part by DARPA, NGA, and NSF.]

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.