Abstract: In order to rapidly determine a most likely code word at a decoder provided in a digital data transmission system using block codes or convolutional codes for the purposes of combating a noisy channel environment, an adaptively controllable threshold is introduced so as to discriminate a largest metric with a small amount of computations. The metric is the measure of closeness of a signal transmitted to the decoder to one of plural code words previously stored in the decoder. A counter counts the number of metrics each exceeding the threshold, and informs a controller of the count result, if the count result is more than one, the controller raises the threshold, on the contrary, if the count result is zero then the controller lowers the threshold, both are performed in an effort to narrow down one metric in excess of the threshold. The threshold thus adaptively controlled is compared with each of plural metrics. If the count result becomes one, it implies that the largest metric has been determined.

Abstract: Sequentially decoding a plurality of symbol sets of an incoming data sequence with less amount of computation in an application wherein paths in a code tree do not occur equiprobably is disclosed. A code tree is previously memorized which comprises a plurality of paths defined by a plurality of sequences of nodes. A pointer generator is provided for generating a pointer that defines a node that specifies a path in the code tree. A plurality of branch metric generators each generates a metric of a branch which forms part of a path and which is to be examined with a corresponding symbol set of the incoming data sequence. Further, a plurality of path metric generators are provided which respectively receive the branch metrics from the plurality of branch metric generators and respectively generate path metrics using the branch metrics.