ica

674 Words2 Pages

Extensive simulations were carried out on Fast ICA and proposed ICA with two speech mixtures each sampled at 16KHz.
5.1 Results of recorded speech mixtures
Speech signal is a supergaussian signal which has positive kurtosis value. Spiky probability density function with heavy tails is the important property of random variables of Supergaussian signals whereas Subgaussian signals have a flat probability density function. Most of the real world signals like engine noise are supergaussian in nature.The experimental results of the recorded speech signals are shown in Fig.4(a) –Fig 4(d).

Fig.4(a) and Fig 4(b) show mixture of two recorded speech signals each sampled at 4.8 KHz .Fig.4(c) and Fig. 4(d) show the extracted independent male and female speech components obtained through the proposed ICA algorithm. Because the defined algorithm should be capable of handling real-world sized instances, the 80000 samples are taken for processing.
5.1 Convergence Analysis
The convergence analysis is done from the simulation results obtained from NCsim Tool v10. Convergence speed represents the time taken for each of the algorithms to reach convergence. It is achieved when a vector w(k) and its updated vector w(k+1) are pointing in same direction. The proposed ICA takes 11 iterations and Fast ICA takes 18 iterations to reach convergence and to separate two recorded speech mixtures. Table 1 clearly depicts the convergence performance in terms of number of iterations.

Table 1 Convergence performance
ALGORITHM Number of Iterations Taken to Recover
1st speech 2nd speech
Fast ICA 11 18
Proposed ICA 11 11
The total computational time is 900ns for FastICA and 300ns for Proposed ICA. The convergence...

... middle of paper ...

... Component Analysis. IEEE Trans. Industrial Electronics, 54: 548-558. DOI: 10.1109/TIE.2006.885491
[20]. Kim, C.M., HM. Park, T. Kim, Y.K. Choi and S.Y. Lee, (2003). FPGA implementation of ICA algorithm for blind signal separation and adaptive noise canceling. IEEE Trans. Neural Netw., 14: 1038-1046. PMID: 18244558
[21]. Kuo-Kai Shyu, Ming-Huan Lee, Yu-Te Wu, and Po-Lei Lee.( 2008) Implementation of Pipelined FastICA on FPGA for Real-Time Blind Source Separation. IEEE Trans. On Neural Networks, Vol. 19, No. 6 PMID18541497
[22]. Dinesh, P., N. Das and A. Routray,( 2011). Implementation of Fast-ICA: A Performance Based Comparison Between Floating Point and Fixed Point DSP Platform. Measurement Sci. Rev., 11: 119-18. http://connection.ebscohost.com/c/articles/69721126/implementation-fast-ica-performance-based-comparison-between-floating-point-fixed-point-dsp-platform

More about ica

Open Document