Leveraging speech-active regions towards active safety in vehicles

Since audio/speech based systems are gaining a lot of prominence in in-vehicular technology, it becomes very important to understand the effect of speech (i.e., speaking or listening) on drivers´ cognitive workload and overall driving performance. Towards this goal, this study probes into the impact of speech on driver distraction, and proposes a method for detecting the active-speech regions to be utilized in driver distraction detection. It is shown that the fusion of speech and vehicle dynamics signals helps in better assessing the distraction level, thus more relevant corrective commands/warnings can be issued by the vehicle safety system. It is observed that while in-vehicle speech by itself does not necessarily cause distraction, it plays a significant role in increasing the distraction level of the driver if he is performing any secondary tasks, and is engaged in a conversation.