Multiuser Channel Estimation in Wavelet Packet-UWB Systems

Centralized maximum likelihood (ML) data aided channel estimation for wavelet packets based ultra wideband (WP-UWB) transmissions is investigated. Training pilot symbols are employed for joint ML channel estimation based on a least square algorithm derived from the optimum ML approach. An optimum training sequence criterion is derived. Based on this, nearly optimum perfect root of unity sequences (PRUS) are tested and compared to Gold sequences and random sequences. In this estimation strategy, a generalized inverse of the training pilot symbols matrix can be pre-computed and stored in memory. Furthermore, PRUS allows to save the inverse matrix operation and so reducing the complexity of the joint ML estimator significantly. This technique is especially attractive for low data rate transmissions as in sensor networks because of its low complex implementation. Performance analysis results in terms of the MSE and BER for a single user matched filter are provided. Trade-offs between performance and complexity are described as well. The investigated channel estimation algorithms for UWB channels offer flexibility, low complexity and robustness, making channel estimation fairly simple and effective.