| This dissertation is part of a sponsored research contract, the objectives of which were (1) to perform a detailed review of the existing scientific literature on human factors and speech recognition technology, and the literature supporting the use of speech input for controlling in-vehicle tasks; (2) to develop a decision tool that would aid designers in determining whether a given in-vehicle task function should be performed using speech input or some method of manual control input; (3) to experimentally examine a number of automatic speech recognition (ASR) system parameters, input modalities, and driver ages to evaluate the effects these variables had on driving performance, task-function usability, and driver preference/acceptance of task-function design; and (4) to develop a comprehensive set of human factors guidelines and recommendations for the use of ASR technology in automotive applications.; The first part of the dissertation is a review of the literature on attentional resource theories, stimulus-central processing-response (S-C-R) compatibility, and mental workload.; Next, a decision tool and methodology are described for evaluating whether a given in-vehicle task function should be performed using speech input or some other manual control input method.; Two experiments were conducted to evaluate the effects of ASR system design parameters, input modality, and age on driving performance, task-function usability, and driver preference/acceptance of task-function design.; Results from the two empirical studies showed that manipulation of the various ASR system design parameters (i.e., recognition error type, recognition accuracy, and input condition) significantly affected measures of driving performance, system usability, and driver preference/acceptance. However, from a practical viewpoint, ASR system design parameters had a nominal effect on driving performance. Differences measured in driving performance brought on by changes in ASR system design parameters were small enough that alternative ASR system designs can be considered without impacting driving performance. The speech-input methods tested were not practically different from current manual-input methods used in performing identical in-vehicle tasks for any of the measures of driving performance recorded. Therefore, no benefits can be claimed for ASR systems improving driving safety/performance compared to current manual-control systems. (Abstract shortened by UMI.)... |
