A new look at the informational gain of soft decisions
Lexa, Michael; Johnson, Don
This paper develops a new systematic method of studying the benefits of 2-bit soft decisions by applying the concepts of information processing theory. We quantify performance in terms of the information transfer ratio and demonstrate the performance gain over hard decision detectors in several noise environments. In addition, we show that likelihood ratio tests maximize the information transfer ratio, and we propose a method of optimizing threshold values for the 2-bit soft decision detector.