Exploiting pitch dynamics for speech spectral estimation using a two-dimensional processing framework

摘要

This thesis addresses the problem of obtaining an accurate spectral representation of speech formant structure when the voicing source exhibits a high fundamental frequency. Our work is inspired by auditory perception and physiological modeling studies implicating the use of temporal changes in speech by humans. Specifically, we develop and evaluate signal processing schemes that exploit temporal change of pitch as a basis for high-pitch formant estimation. As part of our development, we assess the source-filter separation capabilities of several two-dimensional processing schemes that utilize both standard spectrographic and auditory-based time-frequency representations. Our methods show quantitative improvements under certain conditions over representations derived from traditional and homomorphic linear prediction. We conclude by highlighting potential benefits of our framework in the particular application of speaker recognition with preliminary results indicating a performance gender-gap closure on subsets of the TIMIT corpus.