Adaptive Algorithms for Independent Component Analysis: Formulations and Application to CDMA Communication system with Electronic Implementation by

Blind Source Separation (BSS) is a vital unsupervised stochastic area that seeks to separate the underlying source signals from their mixtures with minimal assumptions about the source signals and/or the mixing environment. BSS has been an active area of research and in recent years has been applied to numerous domains including biomedical engineering, image processing, wireless communications, speech enhancement, remote sensing, etc. Most recently, Independent Component Analysis (ICA) has become a vital analytical approach in BSS. In spite of active research in BSS, however, many foundational issues still remain in convergence speed, performance quality and robustness in realistic or adverse environments. Furthermore, some of the developed BSS methods are computationally expensive, sensitive to additive and background noise, and not suitable for a real-time or real world implementation. In this thesis, we first formulate new effective ICA-based measures and their corresponding robust adaptive algorithms for the BSS in dynamic “convolutive mixture” environments. We demonstrate their superior performance to present competing algorithms. Then we tailor their application within wireless (CDMA) communication systems and Acoustic Separation System. We finally explore a system realization of one of the developed algorithms among ASIC or FPGA platforms in terms of cost, effectiveness, and economics of scale. We firstly investigate several measures which are more suitable for extracting different source types from different mixing environments in the learning system. ICA for instantaneous mixtures has been studied here as an introduction to the more realistic convolutive mixture environments. Convolutive mixtures have been investigated in the time/frequency domains and we demonstrate that our approaches succeed in resolving the standing problem of scaling and permutation ambiguities in the presentt research. We propose a new class of divergence measures for Independent Component Analysis (ICA) in order to demix sources. The Convex Cauchy-Schwarz Divergence (CCS-DIV) is formed by integrating convex functions into the Cauchy-Schwarz inequality. The new measure is symmetric and convex, where the degree of convexity can be tuned by a (convexity) parameter. Due the convexity, the CCS-DIV is more likely to attain the optimal minima and speedup the search process. A non-parametric (ICA) algorithm generated from the proposed divergence is developed with different convexity parameters and employing the Parzen window-based distribution. The novel contrast function results in effective parametric and nonparametric ICA algorithms. Moreover, Two pairwise iterative schemes are proposed to tackle the high dimensionality of sources. Two pairwise non-parametric independent component analysis ICA algorithms are introduced based on the new high-performance Convex Cauchy–Schwarz Divergence (CCS-DIV). These two schemes enable fast and efficient de-mixing of sources in real-world applications where the dimensionality of the sources is high. Secondly, the more challenging problem in communication signal processing is to estimate the source signals and their channels in the presence of other co-channel signals and noise without the use of a training set. Blind techniques are promising to integrate and optimize the wireless communication designs i.e. equalizers/ filters/ combiners through its potential to suppress the inter-symbol interference (ISI), adjacent channel interference, co-channel and the multi access interference MAI. Therefore, a new blind detection algorithm, based on fourth order cumulant matrices, is presented and applied to the multi-user symbol estimation problem in Direct Sequence Code Division Multiple Access (DS-CDMA) systems. The blind detection is to estimate multiple symbol sequences in the downlink of a DS-CDMA communication system using only the received wireless data and without any knowledge of the user spreading codes. The proposed algorithm takes advantage of higher cumulant matrix properties to reduce the computational load and enhance performance. In addition, we address the problem of blind multiuser equalization in the wideband CDMA system, in the noisy multipath propagation environment. Herein, we propose three new blind receiver schemes, which are based on the state space structures. This so-called blind state-space receivers (BSSR) does not require knowledge of the propagation parameters or spreading code sequences of the users but relies on the statistical independence assumption among the source signals. We then develop and derive three update-laws in order to enhance the performance of the blind detector. Also, we upgrade three semi-blind adaptive detectors based on the incorporation of the RAKE receiver and the stochastic gradient algorithms which are used in several blind adaptive signal processing algorithms, namely FastICA, RobustICA, and principle component analysis PCA. Through simulation evidence, we verify the significant bit error rate (BER) and computational speed improvements achieved by these algorithms in comparison to other leading algorithms. Lastly, system realization of one of the developed algorithms has been explored among ASIC or FPGA platforms in terms of cost, effectiveness, and economics of scale. Based on our findings of current stat-of-the-art electronics, programmable FPGA designs are deemed to be the most effective technology to be used for ICA hardware implementation. Journal Publications  Zaid Albataineh and Fathi Salem, “A Blind Adaptive CDMA Receiver Based on State Space Structures” IEEE Trans. Wireless communication 2014 (to be Submitted)  Zaid Albataineh and Fathi Salem, “Convex Cauchy-Schwarz Independent Component Analysis CCS-ICA for Blind Source Separation,” IEEE Trans. Neural Netw. 2014(Under-Review)  Zaid Albataineh and Fathi Salem, “Robust Blind Multiuser Detection Ds-CDMA Algorithm Using Simplified Fouth Order Cumulant Matrices,” Circuit, Systems and Signal Processing, Springer—previous version appeared-ICAS2013. (Accepted)  Zaid Albataineh and Fathi Salem, “RobustICA-Based Algorithm for Blind Separation of Convolutive Mixtures” IEEE Trans. Circuit and System I 2014 “to be Submitted)  Zaid Albataineh and Fathi Salem, “High Gain, High Linearity Mixer for 1.9GHz Receiver FrontEnds”, In Proc. of the International Journal of Research in Wireless Systems (IJRWS), Vol. 2. No. 1. pp. 1-6. March 2013.  Zaid Albataineh and Fathi Salem, " An Energy-Efficient and High Gain Low Noise Amplifier for Receiver Front-Ends", In Proc. of the International Journal of Research in Wireless Systems (IJRWS), Vol. 1, Issue 2, pp.24-29, December 2012.  Zaid Albataineh, Fathi Salem and Jehad Ababneh, “Linear Phase FIR Low Pass Filter Design Using Hybrid-Differential Evolution”, In Proc. of the International Journal of Research in Wireless Systems (IJRWS), Vol. 1, Issue 2, pp. 36-45, December 2012. Letters  Zaid Albataineh and Fathi Salem, “A Convex Cauchy-Schwarz Divergence Measure for Blind Source Separation” Springer. Neural Processing Letter 2014 (Submitted)  Zaid Albataineh and Fathi Salem, “Two Pairwise Iterative Schemes For High Dimensional Blind Source Separation,” Springer. Neural Processing Letter 2014(Submitted)