Speech enhancement in transform domain

• Main Author: Ding, Huijun
• Other Authors: Soon Ing Yann
• Format: Theses and Dissertations
• Language: English
• Published: 2011
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
• Online Access: https://hdl.handle.net/10356/43536
• Institution: Nanyang Technological University

This thesis focuses on the development of speech enhancement algorithms in the transform domain. The motivation and objectives of the work are first stated and various effects of noise on speech are discussed. A literature review of various speech enhancement algorithms with an emphasis on those implemented in the transform domain is presented. Some of the important speech enhancement algorithms are outlined and various transform methods are compared and discussed. Since the primary aim is to attenuate the noise component of a noisy speech in order to enhance its quality using transform based filtering algorithms, three proposed transform domain speech enhancement algorithms are introduced in details. The first one is DCT-based algorithm which reduces the variance between frames and achieves a better noise reduction. Next a DFT-based speech algorithm is introduced which considers the 2D relations of DFT coefficients by incorporating image processing type of techniques. The last one is a DFT-based post-processor which is proposed for over-attenuated speech components regeneration. For the purpose of evaluation, several objective measures are utilized together with two novel proposed objective measures. The proposed objective measures are able to give specific judgements on speech distortion and noise reduction. All the proposed algorithms are compared in terms of their computational cost, inherent delay times as well. The strengths and weaknesses of various proposed algorithms are analyzed and recommendations for future work are presented.