Docstoc

Appendix A

Document Sample
Appendix A Powered By Docstoc
					                                                                                        Appendix A




Appendix A

Initial Mean Convergence Derivation

Assuming that in the initial adaptation, e(n) is correlated to X(n) that the principle of

orthogonality is not application at the initial stage (E{e(n)x(n) ≠ 0). Since x(n) and e(n) are
Gaussian distributed, the Gaussian Moment Factoring Theorem is used to simplify the higher
order moment of e3(n) [Haykin 1991 and McGee 1971]:


                             E{V(n+1)} = E{V(n)} + γE{e3(n)X(n)}.                            (A.1)


Using the Gaussian moment factoring theorem:


                             E{e3(n)X(n)} = 2[E{e2(n)}E{e(n)X(n)}].                          (A.2)
Replacing (A.2) to (A.1):


                      E{V(n+1)} = E{V(n)} + 2γ[E{e2(n)}E{X(n)e(n)}],                         (A.3)


Since, e(n) = η(n) - XT(n)V(n),


              E{V(n+1)} = E{V(n)} + 2γ[ E{e2(n)}E{X(n)[η(n) - XT(n)V(n)]}].                  (A.4)


Again using Assumption B, that E{X(n)η(n)} = 0, equation (A.3) is simplified to:


      E{V(n+1)} = E{V(n)} - 2γE{e2(n)}E{X(n)XT(n)}E{V(n)},                                   (A.5)

                                                       |
              E{V(n+1)} = [ I - 2γE{e2(n)R]E{V(n)} ||W(n)||<<||W*||.                         (A.6)



(A.5) represents the initial adaptation of the mean weight error difference equation.




                                              163
                                                                                Appendix A




Appendix B

Publications



        1. M.K. Chan, C.F.N. Cowan, “Normalised Least Mean Fourth (XE-NLMF)
        Algorithm with Faster Convergence,” IEEE International Conference on Acoustics,
        Speech, and Signal Processing ICASSP2001, Student Forum, Salt Lake City (US),
        May 2001.


        2. M.K. Chan, C.F.N. Cowan, “A Normalisation Technique For The Least Mean
        Fourth Algorithm In System Identification,” Irish Signal and System Conference
        ISSC2001, pp. 141-146, Maynooth (Ireland), June 2001.


        3’. M.K. Chan, V. Stewart, C.F.N. Cowan, “A Parallel Adaptation for Speech Echo
        Cancellation,” Irish Signal and System Conference ISSC2002, pp. 201-205, Cork
        (Ireland), June 2002.


        4. M.K. Chan, V. Stewart, C.F.N. Cowan, “Multiple Adaptive Filtering for Speech
        Echo Cancellation,” IEE Non-linear and Non-Gaussian Signal Processing Workshop
        N2SP, Peebles (Scotland), July 2002.


        5’. M.K. Chan, C.F.N. Cowan, “Using A Normalised LMF Algorithm For Channel
        Equalisation With Co-Channel Interference,” XI European Signal Processing
        Conference Eusipco2002, Toulouse (France), Vol. 2, pp. 49-51, September 2002.


        6. M.K. Chan, V. Stewart, C.F.N. Cowan, “"Multiple Adaptive Filtering Scheme For
        Non-Gaussian Speech Signal in Echo Cancellation,” submitted to: IEE Proceedings -
        Vision, Image and Signal Processing, Special issue on ‘Non-linear and Non-Gaussian
        Signal Processing’, 2002.




’ reprinted at back of the thesis.




                                           164
      Appendix A




201

165
      Appendix A




202


166
      Appendix A




203



167
      Appendix A




204



168
      Appendix A




205




169
        Appendix A
      Eusipco 2002




49




170
      Appendix A




50



171
      Appendix A




51



172
      Appendix A




52



173

				
DOCUMENT INFO