Maximum likelihood detector which splits the trellis into sections in time and hierarchically processes it, combining adjacent sections at each level

摘要

The maximum likelihood detector calculates a most likely path across a trellis by initially generating weights indicating the likelihood of all of the various transitions within the trellis from the received data-stream (step 1). The trellis is then split into sections which contain loops, that is sections which have two paths from each initial state to each final state (step 1 broken down). For each loop the most likely path is selected as the survivor path and passed on to the next stage of processing. Adjacent sections are now combined to create new loops and the path selection process is repeated (steps 2 and 3). This step is repeated until a path has been determined for the whole trellis (steps 3 and parallel traceback). The system allows parallel hierarchical operation (see Fig. 5) which leads to faster processing a reduced latency. The system is scalable to more complex trellises (eg E2PR4 see Figs. 7,8) and applicable to hard disk readers and communication receivers.