Abstract: Speech signals are recorded by a distant microphone in an enclosed space, frequently tainted by interferences of many speech signals and reverberations of room. Accordingly, the separation of speech signals is important and allows the process of Blind Source Separation (BSS) and Blind De-reverberation (BD). The signified frequency-domain BSS and Independent Component Analysis (ICA) are utilized for separation process. Subsequently, the permutation ambiguities of the ICA solutions are prearranged and the separated signals are shaped accurately in the time domain. While most speech enhancement algorithms improve the excellence of speech but not increase speech transparency in reverberation. This motivates the improvement of an algorithm that can be modified for an acoustic environment and recover speech transparency. This study proposes the Adaptive Equalization Method for dereverberation and Savitzky-Golay filtering is used to reduce the noise.
Jasmine J.C. Sheeja and B. Sankara Gomathi, 2014. Blind Techniques for Improving Speech Mixtures Using an Adaptive Method. Research Journal of Applied Sciences, 9: 262-273.