An intelligent HF squelch

For many years a requirement has existed for a reliable HF squelch system to switch the output audio of an HF receiver only when speech is present. This paper describes the development of such a system. The speech detection processing is based on a design combining a neural network with algorithms that extract salient features from the audio signal. The emphasis of the work was to produce a real-time system to a given specification, rather than to attempt to investigate all the different possible approaches. It should be appreciated that the speech detection processing is quite different from speech recognition, where individual words are recognised, usually in environments of little noise. Currently, speech recognition is limited in terms of number of words and number of speakers. The speech detection developed for this application is designed to detect all words from all speakers, in a very noisy environment