This item is available under a Creative Commons License for non-commercial use only
Electrical and electronic engineering
Presence of noise in the speech can sometimes become annoying as it can lead to loss of important data or create misunderstandings between the communications area which can lead to major problems associated to loss of time and money. This thesis focuses to filter out noise form a speech signal which is simulated in Matlab/Octave software while making a comparison between temporal resolution of signal with respect to the spectral resolution of the signal in which the parameters such as the size of window length are varied in order to obtain the best speech separation performance. To get the best spectral and temporal resolution with respect to the window length in order to find out the presence of speech sound in the mixture or how strongly the mixture is dominated by the noisy signal. The reconstructed signal is the original speech sound which was applied at the input. To study the relationship between window-disjoint orthogonality and window length and to get the best separation performance.
Maurya, A. (2017) Interference Unmixing and Estimation Technique for Improvement of Speech Separation Performance. Masters thesis, DIT, 2017.