Refined Kalman smoothing for decoding real convolutional codes

Abstract

Real number convolutional codes can be corrected by time-varying smoothed Kalman estimation. The smoothing process can be truncated whenever the tracking innovations fall below a precomputed matrix norm.