Adaptive quantisation of input value range e.g. for classification of patterns - determines relative frequency with which input values are associated with partitions to calculate estimated densities, and divides input values into new partitions so that difference between densities does not exceed limit

摘要

An adaptive process is used for quantifying the input range of a sampled parameter. The system has four microprocessors (1-4), two interfaces (12,13), memories (5,6,8,9,10) and a low-pass filter (11). Data is read into the system and partition boundaries are defined from the values. The partition lines are equidistant from each other. The number of partitions are such that limited variations in input value distributions exist. ADVANTAGE - Improvements in histogram data representations.