On the Lempel-Ziv-Welch coded speech transmission and markov chain detection

In this paper, we present a novel technique for speech transmission over noisy channels using Lempel-Ziv-Welch (LZW) source coding and Markov chain detection. The basic idea is to make use of the highly dynamic nature and high compression ratio of LZW algorithm that is possible for speech signals due to its correlation. However, the erroneous reception after noise is added to this coded signal poses challenge of dynamic dictionary forming. In this paper, assuming that speech maintains high order memory, this has been non-trivially handled by high order Markov chain (MC) modeling, detection and correction technique. This technique ensures that the decoded speech maintains the approximate correlation normally present in speech. In order to justify the quality of speech we have made use of spectrogram analysis. Few simulations for the whole process have been performed to illustrate the performance. This speech compression technique can be used real-time and imposes no outstanding delay at the reception side except the transmission delay plus the minimal decoding and detection time