Abstract:
The article proposes a method of coding sources with finite memory, whose statistics are unknown. Unlike existing methods, it results in reduced redundancy in the case in which the source memory is larger than that of the coder and decoder.