On Codeword Lengths Guaranteeing Synchronization

Vladimir V. Gusev, Elena V. Pribavkina

Abstract: Prefix codes such as Huffman codes are commonly used for loseless data compression. The class of synchronizing codes is often chosen to improve error resilience or to enable parallel decoding of data. Such codes have a special sequence whose occurrence realigns decoding process leading to recovery from errors in a data stream. In the present paper we identify a class of codes whose synchronizability depends only on the lengths of codewords. Namely, we show that every maximal finite prefix code with only two codeword lengths is synchronizing if and only if these lengths are coprime.