Processing of the manual morse signal using optimal linear filtering, smoothing and decoding

Abstract

This thesis investigates the problem of automatic transcription of the Morse signal, and describes and documents several approaches to filtering, processing and decoding it for transcription. The baseband signal is first modeled as a modified random telegraph wave. A discrete Kalman filter and a linear smoother are then used to process the demodulated signal in order to gain a measure of the effectiveness and applicability of this model. It is shown experimentally that this model and processing yield a significant reduction in the transcription error rate. Next a Viterbi decoder algorithm based on a simple Markov model of the code is incorporated in a more general model for pre-detection Kalman filtering. It is shown that this filter permits acceptable recovery of Morse signals whose average signal-to-noise ration is as low as -14 dB in a 3 kHz bandwidth.