Academic journal article Human Factors

Automatic Speech Recognition in Adverse Environments

Academic journal article Human Factors

Automatic Speech Recognition in Adverse Environments

Article excerpt

INTRODUCTION

Automatic speech recognition (ASR) technology allows human speech signals to be used in order to carry out preset activities. Upon detection and recognition of a sound or string of sounds, the recognizer can be programmed to execute a predetermined action. Consequently, many groups of individuals have benefited from using ASR in human-machine interaction, in human-to-human communications, and as a means of control in their immediate environments.

For instance, ASR in industrial applications allows operators to interact with computer systems from remote points on the shop floor and permits users to interact with machines while performing an additional task - for instance, manipulating a parcel or measuring tolerances on an electrical component (Noyes, Baber, & Frankish, 1992). Likewise, ASR has been successfully used by groups of disabled individuals for environmental control, such as wheelchair operation and switching domestic appliances and robotic arms on and off (Noyes & Frankish, 1992).

Although ASR has a number of operational applications and many potential applications still in development, environmental factors play a major role in influencing the successful use of ASR technology. As Simpson, McCauley, Roland, Ruth, and Williges (1985) stated, "the task environment comprises a number of factors that must be studied for their effect on human performance and therefore on speech task design" (p. 123). Moreover, the success or failure of an ASR application can be partly determined by consideration of these environmental factors. In some application domains, a number of factors could have an impact on successful ASR use. For example, high ambient noise levels may lead to the degradation of the speech signal, and elevated levels of stress could alter the way in which operators respond to task demands, including requirements to produce speech. However, the influence of environmental factors can be shaped by human mediation - that is, by decisions and efforts made by the user to counter the possible effects of these factors. Mediation effects are likely to be inconsistent across operators for different environmental factors. A range of other human factors issues - such as system design, training, and personnel selection - will also influence the extent to which environmental factors have a detrimental effect on ASR use.

Considerable research has been aimed at implementing mediation through the speech recognition technology, particularly with respect to ambient noise. Efforts have been directed at building ASR devices that are robust enough to cope with the extremes of vibration, pressure, and acceleration, as found in the aircraft cockpit environment. Furthermore, several ASR companies are now marketing devices that have been tailored for specific application environments - for instance, the Interactive Voice Module developed by Smiths Industries for military cockpits. Some human factors engineering effort has focused on reducing the direct effects of environmental factors on human performance with ASR, such as noise and acceleration (Taylor, 1986). However, relatively little attention has been given to the potential effects of indirect environmental factors, such as stress and workload, on ASR use. Although mediation can affect system efficiency, it is unlikely that mediation will be consistent for all operators across all environmental conditions. Not only will different environmental conditions require different forms of mediation, but there will also be tremendous individual differences among operators in terms of coping strategies and other forms of mediation.

A report prepared for the National Research Council (NRC) in the United States by the Committee on Computerized Speech Recognition Technologies considered the potential problems arising from the use of ASR in environments characterized by psychological and physiological stress (NRC, 1984). The authors of the report concluded that ASR was "not sufficiently advanced to provide robust, reliable performance in hostile and high stress environments" but that it could support "restricted applications in benign environments" (p. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.