Complexity reduction in subspace-based blind channel identification for DS/CDMA systems

Direct-sequence code-division multiple access is emerging as a potential multiple-access communication scheme for future digital wireless communications systems. Such wide-band systems usually operate in a frequency-selective fading channel that introduces intersymbol interference and thus potential performance degradation. Previously proposed subspace-based blind channel identification algorithms, which provide estimates of channel parameters for effective equalization, suffer from high numerical complexity for systems with large spreading gains. In this paper, it is shown that, through the use of matched filter outputs, reduction in numerical complexity can be obtained. The complexity reduction is considerable when the channel length is small and the system is moderately loaded. The results show that the new algorithm suffers a slight performance loss. Although the employed matched filter outputs do not form a set of sufficient statistics for the unknown channels, the difference between the matched filter outputs and the sufficient statistics becomes negligible for large observation lengths and the asymptotic normalized Fisher information does not change. Performance is evaluated through simulations, the derivation of a tight approximation of the mean-squared channel estimation error, and through comparisons to the Cramer-Rao bound for the estimation error variance. It is shown that the approximation of the mean-squared error can be obtained in terms of the correlation of the spreading codes and the channels. This representation of the error supplies a tool for investigating the relationship between performance and spreading sequence correlations