Blind deconvolution in spread spectrum communications over non minimum phase channels

Higher than second order statistics is a powerful signal processing tool and a lot of research effort has been made to find new and implementable cumulants based algorithms for system identification and deconvolution of non minimum phase systems for the application in cases of practical interest. In communications problems equalization in the presence of time dispersive channels (usually non minimum phase) is still discussed since traditional autocorrelation (i.e. second order statistics) based methods cannot be used. The authors propose, stemming from the idea of Shalvi and Weinstein (1993), a method for blind channel estimation and symbol detection which seems to be particularly interesting for its fast convergence and conceptual simplicity. The despreading operation is performed jointly with channel distorsion compensation using the algorithm described in Shalvi and Weinstein. The couple correlator/channel is in fact interpreted as a unique FIR (finite impulse response) system whose taps are given by the convolution of the spreading code and the impulse response of the channel. The problem is so reduced to an autoregressive (AR) model to be estimated. Then the receiver is simply constituted by a tapped delay line which is the inverse of the AR model. The channel model is frequency selective slowly fading typical of the L-band satellite channel. The method is faster than traditional methods also over minimum phase channels but there is an increased computational complexity