Studies in Signal Processing Techniques for Speech Enhancement: A comparative study

Shivamurti, M and Narasimhan, SV Studies in Signal Processing Techniques for Speech Enhancement: A comparative study. Other. NAL, Bangalore. (Unpublished)

Full text available as:
[img] PDF
Download (703Kb)

    Abstract

    Speech enhancement is very essential to suppress the background noise and to increase speech intelligibility and reduce fatigue in hearing. There exist many simple speech enhancement algorithms like spectral subtraction to complex algorithms like Bayesian Magnitude estimators based on Minimum Mean Square Error (MMSE) and its variants. A continuous research is going and new algorithms are emerging to enhance speech signal recorded in the background of environment such as industries, vehicles and aircraft cockpit. In aviation industries speech enhancement plays a vital role to bring crucial information from pilot’s conversation in case of an incident or accident by suppressing engine and other cockpit instrument noises. In this work proposed is a new approach to speech enhancement making use harmonic wavelet transform and Bayesian estimators. The performance indicators, SNR and listening confirms to the fact that newly modified algorithms using harmonic wavelet transform indeed show better results than currently existing methods. Further, the Harmonic Wavelet Transform is computationally efficient and simple to implement due to its inbuilt decimation-interpolation operations compared to those of filter-bank approach to realize sub-bands.

    Item Type: Proj.Doc/Technical Report (Other)
    Subjects: ENGINEERING > Electronics and Electrical Engineering
    Division/Department: Aerospace Electronics and Controls Division, Aerospace Electronics and Controls Division
    Depositing User: Mr M shivamurti
    Date Deposited: 23 Jun 2011 10:02
    Last Modified: 23 Jun 2011 10:02
    URI: http://nal-ir.nal.res.in/id/eprint/9241

    Actions (login required)

    View Item