Wiener Filtering in Frequency Domain to Enhance Speech Corrupted by Colored Noise
V. Sailaja1, P. Sunitha2, B. Vasantha Lakshmi3, V. Prasanth4
1Dr. V. Sailaja, Professor, Pragati Engineering College, Surampalem (Andhra Pradesh), India.
2P. Sunitha, Associate Professor, Pragati Engineering College, Surampalem (Andhra Pradesh), India.
3B. Vasantha Lakshmi, Associate Professor, Pragati Engineering College, Surampalem (Andhra Pradesh), India.
4V. Prasanth, Associate Professor, Pragati Engineering College, Surampalem (Andhra Pradesh), India.
Manuscript received on 13 October 2019 | Revised Manuscript received on 22 October 2019 | Manuscript Published on 02 November 2019 | PP: 1058-1062 | Volume-8 Issue-2S11 September 2019 | Retrieval Number: B11790982S1119/2019©BEIESP | DOI: 10.35940/ijrte.B1179.0982S1119
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This paper presents a method for speech enhancement to predict speech quality in presence of highly non-stationary scenarios using basic wiener filtering in frequency domain with an adaptive gain function under eight different noises at three different ranges of input SNR. Its performance is evaluated in terms of objective quality measures like LPC based spectral distortion measures are Cepstrum Distance, Itakura Saito and Log Likelihood Ratio. This method was tested using Noizeous database, its performance measures were compared against spectral subtractive type algorithms and it shows its improvements in terms of objective quality measures.
Keywords: Speech Enhancement, Wiener Filtering, Spectral Subtraction, Noise, Objective Quality Measures.
Scope of the Article: Frequency Selective Surface