Use this URL to cite or link to this record in EThOS:
Title: Modulation-domain Kalman filtering for single-channel speech enhancement, denoising and dereverberation
Author: Dionelis, Nikolaos
ISNI:       0000 0004 8499 4793
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis is about robust single-channel speech enhancement, joint noise suppression and dereverberation, using modulation-domain Kalman filtering to blindly and adaptively track the time-frequency log-magnitude spectra of speech and reverberation. The focus of this work is the development of speech enhancement algorithms which operate on the log-spectral domain representation of single-channel degraded (noise and reverberation) speech recordings. The Kalman filtering framework is used extensively for algorithm development and the algorithms developed in this thesis are relevant to researchers interested in single-channel speech enhancement. In this work, several adaptive phase-sensitive speech enhancement algorithms based on modulation domain Kalman filtering are designed, described, implemented and evaluated that perform blind joint denoising and dereverberation, while accounting for the inter-frame speech dynamics by estimating the posterior distribution of the speech log-magnitude spectrum given the log-magnitude spectrum of the noisy reverberant speech. The algorithms are composed of different processing blocks and techniques; understanding the implementation choices that are made during the system design is important as this provides insights that can assist the development of new speech enhancement systems for both noise reduction and dereverberation. The Kalman filter update step models the non-linear relations between the speech, noise and reverberation log-spectra. The Kalman filtering algorithm uses a signal model that takes into account and tracks the reverberation parameters of the reverberation time, T60, and the direct-to-reverberant energy ratio (DRR) to improve the estimation of the speech log-magnitude spectrum. Phase-sensitive modulation-domain Kalman filtering in the log-spectral domain can now be performed and can now be better understood with this research work. The Kalman-filter-based enhancement algorithms, dependent on the signal model and the tracked quantities, are evaluated in terms of speech quality, speech intelligibility and dereverberation performance for a range of reverberation parameters and SNRs, in different noise types and reverberant conditions, and are also compared to competing denoising and dereverberation techniques. Experimental results and instrumental measures indicate that the proposed algorithms enhance the speech quality of the degraded noisy and reverberant signals and outperform other algorithms over a range of SNRs for various noise types and reverberant conditions.
Supervisor: Brookes, Mike Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral