Computer Utilization of Sequential Hypothesis Testing for Detection and Classification of Sonar Signals.

摘要

Active sonar systems furnish the operator large quantities of data to examine for possible targets. In order to detect small target returns, the display thresholds must be set such that clutter is significant. On a PPI display this clutter makes it difficult to visually integrate successive pings. This report shows a technique utilizing a digital computer to examine the data out of a sonar system, reduce its volume, perform ping-to-ping integration, make simple decisions, and generate useful displays. The process takes advantage of a statistical decision procedure called sequential hypothesis testing. Specifically, the technique is not dependent on a particular sonar system; it is well founded mathematically, and sufficiently simple to make real time implementation possible on a modest state-of-the-art computer. There is no fundamental limit on the number of echo cycles to be considered for ping-to-ping integration, in the decision process. (Author)