More efficient ASIC implementation of digital mobile radio codecs

A novel approach to achieve memory savings in maximum likelihood decoding Viterbi decoders is proposed. Based on tracking trellis survivor paths in the decoding decision process using a backward labels technique, rather than the traditional forward labels technique, and exploiting a `shift register´ property of the trellis decoder, memory savings are achievable without loss of decoding performance. Typically in (n,l,m) convolutional codes savings of the order of 20% of memory requirements are possible. The property is presented as a new theorem, including proof for the general (n,k,m) and popular (n,l,m) convolutional codes.