Date of Award

Summer 2002

Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)


Electrical and Computer Engineering

First Advisor

Alexander Haimovich

Second Advisor

Yeheskel Bar-Ness

Third Advisor

Eliza Zoi-Heleni Michalopoulou

Fourth Advisor

Ali Abdi

Fifth Advisor

Javier Garcia-Frias


Reliable detection of signals transmitted over a wireless communication channel requires knowledge of the channel estimate. In this work, the application of expectationmaximization (EM) algorithm to estimation of unknown channel and detection of space-time coded modulation (STCM) signals is investigated. An STCM communication system is presented which includes symbol interleaving at the transmitter and iterative EM-based soft-output decoding at the receiver. The channel and signal model are introduced with a quasi-static and time-varying Rayleigh fading channels considered to evaluate the performance of the communication system. Performance of the system employing Kalman filter with per-survivor processing to do the channel estimation and Viterbi algorithm for sequence detection is used as a reference.

The first approach to apply the EM algorithm to channel estimation presents a design of an online receiver with sliding data window. Next, a block-processing EM-based iterative receiver is presented which utilizes soft values of a posteriori probabilities (APP) with maximum a posteriori probability (MAP) as the criterion of optimality in both: detection and channel estimation stages (APP-EM receiver). In addition, a symbol interleaver is introduced at the transmitter which has a great desirable impact on system performance. First, it eliminates error propagation between the detection and channel estimation stages in the receiver EM loop. Second, the interleaver increases the diversity advantage to combat deep fades of a fast fading channel.

In the first basic version of the APP-EM iterative receiver, it is assumed that noise variance at the receiver input is known. Then a modified version of the receiver is presented where such assumption is not made. In addition to sequence detection and channel estimation, the EM iteration loop includes the estimation of unknown additive white Gaussian noise variance.

Finally, different properties of the APP-EM iterative receiver are investigated including the effects of training sequence length on system performance, interleaver and channel correlation length effects and the performance of the system at different Rayleigh channel fading rates.