Speech Enhancement using Adaptive Filtering with Different Window Functions and Overlapping Sizes

Main Article Content

Senthamizh Selvi R, et. al.


Speech is the essential form of human communication. Speech processing is the research on speech signals and its processing methods. Noise is the unwanted sound in speech. Noise can cause communication issues, hearing problems, psychological health problems and many more. In most of the modern communication systems, speech enhancement plays a vital role. While transmitting the speech signal, the quality of that signal will degrade due to interference in the surrounding it is passing through. This paper is focused on performance analysis on enhancing the speech by using various windows, transformation techniques and overlapping percentage. The Kalman filter is used for filtering degraded speech signal. The windows used to perform analysis on this work are Hanning, Blackman, Hamming and Cosh window. The transformations used are Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT). In this work, the noisy input signal is divided into multiple number of frames, each frames are sent through the window, after which the overlapping of those frames will be done. This overlapped signal will be applied to transformation technique and filtering process will be done. The output signal will be the enhanced and noise will be reduced to a certain extent. In this way, various combinations of windows, transformations and overlapping percentage, the amount of enhancement obtained is measured by taking the values of Signal to Noise Ratio (SNR) and performance analysis is done with those values. The highest value of SNR of 44.2409 dB is obtained by using the combination of Cosh window in FFT transform of overlapping percentage of 50%.


Download data is not yet available.


Metrics Loading ...

Article Details

How to Cite
et. al., S. S. R. . (2021). Speech Enhancement using Adaptive Filtering with Different Window Functions and Overlapping Sizes. Turkish Journal of Computer and Mathematics Education (TURCOMAT), 12(13), 1886–1894. https://doi.org/10.17762/turcomat.v12i13.8841
Research Articles