Decoding of convolutional codes by Viterbi blind algorithm

Abstract

For decoding of convolutional codes by the Viterbi algorithm, the knowledge of channel parameters on non-Gaussian channels is either essential or preferable. When the considered channel is fast time-varying, reliable and robust estimation of the channel parameters by the conventional decision-directed adaptive algorithms is hardly possible. In the present paper, the Viterbi algorithm is applied in a modified way. Firstly a new trellis diagram is constructed at the receiver side according to modified definitions of trellis state and branch. Secondly a new branch metric is defined so that its expression does not explicitly depend on the channel parameters. With these modifications, decoding can be achieved without the knowledge of channel parameters. This method provides a robust alternative for decoding of convolutional codes on fast fading (especially frequency-selective) channels where the estimation of channel parameters becomes more and more difficult.<>