Model selection-based blind adaptive channel estimation algorithms for parsimonious receivers in CDMA systems

We consider a blind channel estimation and a model selection algorithms for the estimation and identification of significant channel parameters in CDMA systems. In conventional estimation algorithms, channel parameters are usually estimated using a channel length that is related to the maximum delay spread of the channel. However, this can result in estimating too many parameters where many of them are either zero or insignificant. In this work we propose algorithms that estimate then identify only significant channel parameters, leading to parsimonious receivers. First, we use userspsila spreading codes and the received signal to form a data model for the system. Then, a blind adaptive channel estimation algorithm is derived to estimate the channel response. This is followed by a model selection algorithm to identify significant channel parameters among the estimated ones. Compared with conventional algorithms, our algorithm: (i) estimates the channel response with a faster convergence rate (ii) has lower estimation errors (iii) identifies significant parameters with high probabilities and at low SNRs (iv) has a lower BER performance.