We propose to develop a statistical framework for nonlinear signal processing for communications by using a recent extension of maximum likelihood estimation, the maximum partial likelihood (MPL) theory. MPL allows for dependent and missing observations and sequential processing of data using only the information that is available at the time of processing. The inclusion of the dependent data in the framework allows development of techniques for joint detection and estimation in the presence of sources and channels with memory, for example to combine maximum likelihood sequence detection with adaptive equalization. The proposed research includes development of a new class of real-time adaptive signal processing algorithms based on MPL estimation by using both gradient optimization and information-theoretic alternating projections, study of their statistical and dynamic properties, incorporation of order/complexity determination into the scheme, and implementations in equalization, joint equalization and sequence estimation, and variable rate speech coding. The educatimn goals of the PI are shaped by the commitment to equip the students with the necessary theoretical and practical facilities to solve real-world problems; to help them mature into resourceful and creative engineers who could meet the challenges of a discipline rapidly growing in complexity. Information on projects in this area and on maximum partial likelihood research can be found at http://engr.umbc.edu/~adali.
