Home // International Journal On Advances in Telecommunications, volume 7, numbers 1 and 2, 2014 // View article

A VAD/VOX Algorithm for Amateur Radio Applications

Authors:
William Forfang
Eduardo Gonzalez
Stan McClellan
Vishu Viswanathan

Keywords: voice activity detection; VAD; voice-activated switch; voice-activated transmission; VOX

Abstract:
Abstract—In amateur radio applications, voice activity detection (VAD) algorithms enable hands-free, voice-operated transmissions (VOX). In this paper, we first review a recent hybrid VAD algorithm, which was developed by combining features from two legacy speech detection algorithms long used in amateur radio applications. We then propose a novel VAD algorithm whose operating principles are not restricted to those of legacy approaches. The new method employs two key features. The first feature, called sub-band variance ratio, is the ratio of energies calculated over a low-frequency region and over the rest of the spectrum of the input audio signal. The second feature, called temporal formant density, is a running N-frame sum of the number of low-bandwidth formants over a low-frequency region. Both features are shown to yield low values for non-speech segments and relatively high values for speech segments. A two-state decision logic that uses these two features is employed to make frame-by-frame VAD decisions, which are then used in the VOX function for amateur radio transmissions. The proposed new method is compared against the hybrid method using both a simple objective measure involving comparisons against manually derived true VAD data and a subjective pairwise comparison listening test, over audio signal data from amateur radio transmissions at various signal-to-noise ratios. The results from these comparison tests show that the new method provides a better overall performance than the hybrid method. In summary, a new VAD/VOX algorithm for amateur radio applications is proposed that offers performance benefits over existing methods.

Pages: 34 to 44

Copyright: Copyright (c) to authors, 2014. Used with permission.

Publication date: June 30, 2014

Published in: journal

ISSN: 1942-2601