Abstract

StartleCam is a wearable video camera, computer, and sensing system, which enables the camera to be controlled via both conscious and preconscious events involving the wearer. Traditionally, a wearer consciously hits record on the video camera, or runs a computer script to trigger the camera according to some pre-specified frequency. The system described here offers an additional option: images are saved by the system when it detects certain events of supposed interest to the wearer. The implementation described here aims to capture events that are likely to get the user's attention and to be remembered. Attention and memory are highly correlated with what psychologists call arousal level, and the latter is often signaled by skin conductivity changes; consequently, StartleCam monitors the wearer's skin conductivity. StartleCam looks for patterns indicative of a ``startle response'' in the skin conductivity signal. When this response is detected, a buffer of digital images, recently captured by the wearer's digital camera, is downloaded and optionally transmitted wirelessly to a webserver. This selective storage of digital images creates a ``flashbulb'' memory archive for the wearable which aims to mimic the wearer's own selective memory response. Using a startle detection filter, the StartleCam system has been demonstrated to work on several wearers in both indoor and outdoor ambulatory environments.