Automatic Speech Recognition, Noise and Workload

Abstract

Despite the increasing use of technology in the developed world, most computer communications still take place via a QWERTY keyboard and a mouse. The use of Automatic Speech Recognition (ASR) whereby individuals can ‘talk’ to their computers has yet to be realised to any great extent. This is despite the benefits relating to greater efficiency, use in adverse environments and in the ‘hands-eyes busy’ situation. There are now affordable ASR products in the marketplace, and many people are able to buy these products and try ASR for themselves. However, anecdotal reports suggest that these same people will use ASR for a few days or weeks and then revert to conventional interaction techniques; only a hardy few appear to persist long enough to reap the benefits. Thus, it is our contention that ASR is a commercially viable technology but that it still requires further development to make a significant contribution to usability. Admittedly, there are some very successful applications that have used ASR for a number of decades, but these are often characterised by relatively small vocabularies, dedicated users and non-threatening situations; typical applications are in offices (Noyes & Frankish, 1989) or for disabled users (Noyes & Frankish, 1992). Given that Armoured Fighting Vehicles (AFVs) could employ ASR with limited vocabulary and dedicated users, the use of ASR in this application is considered here. The principle difference between ASR for AFV and previous applications is the environmental conditions in which the technology will be used.