NONPARAMETRIC DETECTORS WITH APPLICATIONS TO SPREAD-SPECTRUM

DAVID PAUL GRYBOS, Purdue University

Abstract

In general, a communication system is designed by using convenient statistical models for the environmental conditions that effect the system. For many communication systems it is difficult to exactly model these noise and channel characteristics. Thus the sensitivity of a system to variations in the operating environment is often of interest to completely characterize the system performance. A system may have to be designed with little or no information about the environmental characteristics. The performance of such systems under varying conditions is also of interest. Detectors based on rank statistics require only the broadest assumptions about the noise. Locally-optimal-rank detectors are derived for the case of a randomly fading signal corrupted by additive, independent noise. Asymptotic performance results are derived and related to results for the traditional constant signal case. Small-sample performance results are obtained by computer simulation. The case of locally-optimal-rank detectors for zero mean stochastic signals is also briefly considered. In some cases, the problem of designing, and examining the relative sensitivity of a system may not be totally nonparametric. The judicious selection of a variety of statistical models may give a more complete picture of the performance of the system. The urban-mobile spread-spectrum system examined here presents just such a case. A brief overview of the complete land-mobile spread-spectrum system is presented. The remainder of the research involves the design and performance of the receiver used in the proposed system. A convenient model for the proposed receiver is described. The implementation of various receiver structures is then discussed and a possible structure that can be constructed with current technology proposed. Models for the urban-mobile channel are explained and the performance of the receiver in these channels computed. It is shown that the receiver is reasonably robust for a variety of channel models that correspond to average conditions. The performance of the receiver in channels that may possibly correspond to difficult locations in an urban environment is also shown. These performance results will aid in the design of a practical system. The problem of synchronizing the proposed spread-spectrum receiver is briefly considered. It is shown that a simple, continuous-time statistic appears to provide acceptable performance. Synchronization statistics based on sample values and sample ranks are considered and are seen to have sufficiently inferior performance compared to the continuous-time statistic as to make them unlikely candidates for implementation.

Degree

Ph.D.

Subject Area

Electrical engineering

COinS