Optimization and Performance of Multilevel Quantization in Automatic Detectors

Abstract

The optimization and performance of the double-threshold or binary method of automatic radar detection has been the subject of numerous studies in the past, and the loss due to this approximation has been shown to be 1 to 2-1/2 dB. By quantization of the input video into more than two levels, this loss can be avoided; however, little information is available on appropriate design procedures and expected performance. In this paper empirical procedures are described to find good settings of the first and second threshold levels, and the corresponding detection performance is determined. These results, which are believed to be close to optimum, show that for the case of quantization into four levels (2 bits), the loss in decibels is reduced to approximately 1/3 of the corresponding value for the binary case. In some cases this is equivalent to a saving of as much as 1-1/2 dB in signal-to-noise ratio.