Blind Sources Separation Based on Short-Time Discrete Cosine Transforms
Zeng Yanni1,2， ZhangYujie1， QI Rui1,3
COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(12C) 346-353
1School of Mathematics and Physics, China University of Geosciences, Wuhan, China
2 Faculty of Statistics and Applied Mathematics, Hubei University of Economics, Wuhan, China
3School of Science, Naval University of Engineering, Wuhan , PR China
This paper presents a sparse blind source separation method which uses short-time discrete cosine transform (STDCT) to obtain the transformed domain information from a set of linear instantaneous mixtures of these sources. Unlike short-time Fourier transform (STFT) determining the single source point or area by using the ratio of the real and imaginary parts, we remove these points which are away from the mean direction of the cluster using the orthogonal distance between the point and the line. For the clustering part, we use, in this paper, an algorithm inspired from K-means. The final algorithm is easy to extend for any number of sources. Because the STDCT is a Fourier-related transform similar to the STFT, which using only real numbers, so it reduces the computer cost on clustering and improves the algorithm accuracy. Experimental results are provided to evaluate the performance of the proposed algorithm through comparing with STFT from the normalized mean-square error (NMSE) and signal-to-noise ratio (SNR).