Multipath channel deconvolution and variable bit-rate methods for spread spectrum communication systems

This dissertation addresses the signal overlap phenomenon in wireless communication systems. Techniques are proposed for resolving signal overlap in Doppler and zero-Doppler multipath channels, and for taking advantage of signal overlap as a novel tool for improving the performance of wireless networks.; A least squares (LS) method for estimating the parameters of specular Doppler multipath channels is proposed. It outperforms the recently proposed Projection Onto Convex Sets (POCS) algorithm both in execution speed and in estimates' quality. Problems which arise due to phase differences in previously introduced (zero-Doppler) multipath estimation procedures are addressed and solutions to overcome them are suggested. The problem of simultaneous estimation of phase and Doppler is also discussed.; Subtractive Deconvolution (SD), a new method for deconvolution in the time domain, is proposed and analyzed. Its major advantages are its speed, simplicity of implementation, and suitability for real-time processing, while its disadvantage is high sensitivity to noise. The computational load of SD is shown to be less than 1% of that of POCS and LS.; A Genetic Algorithm (GA) is proposed as a viable method for channel deconvolution. It can optimize complex error functions which are often found to be unsolvable by traditional analytical approaches. A novel error function is proposed which emphasizes the minimization of error contributions at delays where the MF output is stronger. GA converges faster when minimizing this error function instead of the classical square norm. In terms of computational load, GA is more complex than POCS, LS, and SD; however, it is also highly parallel in structure, opening up possibilities for achieving competitive speed of execution when implemented on custom-built parallel processors.; Signal overlap can be used to gain an additional degree of freedom in communication network design. An original method for varying the link-rates in a frequency hopping code division multiple access (FH-CDMA) wireless network is proposed. A direct relationship is established between the link-rate and the auto- and cross-interference levels in the receiver's matrix. Using this relationship, FH-CDMA wireless networks can be adaptively optimized to better utilize the time-frequency resources as the traffic fluctuates with time.