Pitch is an important speech parameter in speakerrecognition, speech synthesis, coding, and articulation trainingfor the deaf. The pitch estimation is to determine thefundamental frequency (F0) or period (T0) of a vocal cordvibration causing periodicity in the speech signal. This taskbecomes very difficult when the speech observations areheavily corrupted by noise. Most of the methods proposed inthe literature are capable of estimating pitch from clean speech[1]. As noise obscures the periodic structure of speech, manyexisting methods fail to provide accurate pitch estimates undernoisy conditions.
展开▼
