Algorithm for a memory-based Viterbi decoder

Abstract

A method of convolutional decoding with a memory-based Viterbi decoder employs the property of a trace-back path; that is, the similarity between two consecutive trace-back paths becomes higher as the data error rate goes down. Therefore, the method of the invention saves the previous trace-back path into a register, and as soon as the current trace-back path is found to be the same as the previous one, the demanded path is obtained. After that, the memory read operations will stop, thereby reducing the power consumption caused by memory read operations. Prior to path trace-back, the path prediction can be executed by utilizing the property that the minimum path metric and the path are consecutive. The invention reduces the number of memory access operations and power consumption by employing the mechanisms of path matching and path prediction.