|Date added||June 27, 2004|
|Category||2004 CMRSC XIV Ottawa|
|Tags||Dr. Charles H. Miller Award Winner, Session 5B|
|Author/Auteur||S. Lalande, J.L. Harbluk|
|Award/Prix||Dr. Charles H. Miller|
Speech-based interfaces have been proposed as safer alternatives to visual displays and manual controls in vehicles. The primary advantage of voice input/output in these systems is that they allow drivers to “keep their eyes on the road and their hands on the wheel”. Concerns have been raised, however, that although these interfaces do not require manual interaction or visual attention, they have the potential to distract drivers and degrade safety (e.g., Lee, Caven, Haake and Brown).
The goal of the research reported in this paper was to assess the impact of voice output systems on drivers’ ability to detect events in the driving environment, maintain their driving performance and respond accurately to the information presented in the content of the email messages. Twelve drivers listened to recorded email messages presented in three different audio formats, a human voice and two types of synthetic speech (concatenative and formant), while they drove a route in a driving simulator. Email messages were presented at two levels of difficulty and required yes/no responses from the drivers. In order to assess cognitive distraction, drivers performed an “event detection” task where they had to indicate if a symbol presented in their side mirrors had changed shape. Drivers also drove the route when no email messages were presented to enable a comparison with baseline performance. Drivers also provided ratings of mental effort.
The results revealed an impact of the speech output systems on several of the outcome measures. The results of the event detection task indicated that the type/quality of the speech output system affected the drivers’ accuracy in detecting changes in the driving environment. Specifically, drivers were poorer at detecting these events when either of the synthetic speech systems was used.
There were no observable influences of the speech system type or email message difficulty on the driving performance measures recorded by the driving simulator.
The type of speech system also influenced the drivers’ response accuracy to the messages in the emails. Drivers were less accurate when responding to messages presented in synthetic speech (concatenative) compared with the messages recorded in human voice. Subjective ratings made by the drivers indicated that listening to the synthetic speech required more mental effort than listening to the recorded human voice.
Speech-based systems for in-vehicle use provide an appealing alternative to systems that require visual and/or manual attention from drivers. Proponents of these systems, however, have not always considered the cognitive costs associated with the use of speech-based systems. The results from the present research indicate that although speech-based interfaces are hands-free, using them while driving can negatively impact drivers’ event detection and their performance on the messages themselves. The interface design and safety implications for speech-based in-vehicle technologies are discussed.
S. Lalande, J.L. Harbluk