Speech Recognition and Automotive Applications: Using Speech to Perform in-Vehicle Tasks

Abstract

An experiment was conducted to investigate the effects of automatic speech recognition (ASR) system design, driver input-modality, and driver age on driving performance during in-vehicle task execution and in-vehicle task usability. Results showed that ASR system design (i.e., recognition accuracy and recognition error type) and driver input-modality (i.e., manual or speech) significantly affected certain dependent measures. However, the differences found were small, suggesting that less than ideal ASR system design/performance can be considered for use in automobiles without substantially improving or degrading driving performance. Several of the speech-input conditions tested were statistically similar, as determined by the dependent measures, to current manual-input methods used to perform identical in-vehicle tasks. Further research is warranted to determine how extended exposure to, and use of, ASR systems affects driving performance, in-vehicle task usability, and driver opinion compared with conventional manual-input methods. In addition, the research should investigate whether prolonged exposure to, and use of, ASR systems results in significant improvements compared to the current research findings.