In many applications, the training data to be processed by an adaptive linear estimator can be assumed to have a finite correlation length. An exact analysis for this class of problems that yields the coefficient bias, coefficient correlation matrix, and mean square estimation error is obtained via a stochastic imbedding procedure. A power series expansion in the gain parameter is used to obtain simplified expressions of order one for the above statistical moments. These new expressions are shown to contain the terms that would result from an analysis based upon the assumption of independent training data plus additional terms arising from data correlation. Algorithm convergence properties are studied by identifying the appropriate matrix eigenvalues from the first-order theory.

This paper investigates the use of multilevel differential phase-shift keying with forward error-correcting coding as a technique for achieving limited bandwidth, yet reliable communication over a class of fading channels. Error rate bounds based on the cutoff rate are computed to show that multilevel modulation can be used to allow significant coding gains, while maintaining or reducing the bandwidth of the transmitted signal.

In this note a procedure is given for the numerical evaluation of theM/M/1queue transient state occupancy probabilities which arise in the analysis of dynamic buffer behavior in store and forward networks. The procedure uses the circular coverage function of radar and communication theory to eliminate an infinite series of modified Bessel functions. Computational savings are illustrated by several numerical examples.

The problem of regulating inputs into a tandem store-and-forward node pair with finite queue limits in order to maximize the average message completion rate is considered from the viewpoint of Markov Decision Theory. A variation of the policy iteration procedure due to Howard [3] is used to compute optimum control policies and message completion rates for various configurations of system parameters. Comparative data are provided on the performance of the optimum control law relative to simpler local control laws which require no information sharing.

New results are presented for symbol-by-symbol detection with decision-directed tracking of colored channel disturbances. Recursive sampled-data algorithms are shown for Maximum A Posteriori Probability of detection under colored additive and multiplicative Gaussian noises along with white Gaussian noise. Preliminary evaluation of the algorithms via Monte Carlo simulation shows good performance compared to standard white-noise only algorithms.

A new method of speech digitization called residual encoding is introduced, and its application to the speech digitization problem is studied. The residual encoding system is a form of differential pulse code modulation which utilizes both an adaptive quantizer and an adaptive predictor. The residual encoder differs from previous systems in two ways. First, a sequential estimation method is used to continuously update the predictor coefficients, and second, the predictor coefficients are not transmitted, but are extracted from the estimate of the speech signal at both the transmitter and receiver. No form of pitch extraction is employed. The residual encoding system with a Kalman filter or a stochastic approximation algorithm for identifying the predictor coefficients has produced good quality speech at a data rate of 16 kbit/s.

This paper explores some new solutions to the problem of high-speed data reconstruction in real time. Two types of nonresetting data reconstructors are developed. Both types are simpler than the classical extrapolative hold circuits and when optimized give better performance. The optimization procedure is given and results of simulation studies are included.