Reducing False Positive Rate of Single-Switch Scanning via Patterns in Physiological and Physical Signals?
Brian Leung, Bloorview Research Institute, IBBME University of Toronto and Tom Chau, Bloorview Research Institute, IBBME University of Toronto
Research objectives: To investigate the possibility of a false positive suppression strategy using physiological (skin conductance, skin temperature, blood volume pulse, and respiration) and physical signals (tri-axial accelerometry of the forearms), by estimating the likelihood of switch use during single-switch scanning.
Target population: Single-switch users of all ages. Typically, these individuals are affected by severe and multiple physical disabilities.
Background: Some individuals with severe physical disabilities can only operate one switch due to limited availability of extant physical and physiological pathways for physical access. Moreover, scanning remains a popular paradigm for users to perform non-trivial functional activities with a single switch (e.g., typing with a row-column scanning keyboard). A false positive error is switch activation without a deliberate and correct user action. Correcting a false positive is cumbersome because the user must first undo the unintentional change (if it is reversible) and then reattempt the originally intended action. False positive rate is a function of the switch design and user ability. However, this rate could be lowered if we can determine the likelihood of switch use during scanning and then temporarily disable the switch during periods of low likelihood.
Methods: For data collection (phase 1), 8–10 single-switch users will complete 20 sessions of a single-switch scanning activity (e.g., alphabet matching) while wearing the physiological sensors and accelerometers. The recorded signals will be analyzed off-line and per-participant to determine whether or not a false positive suppression strategy can be developed. Some of the participants will be invited back for phase 2, where we will evaluate the real-time utility of the false positive suppression strategy.
Results:This research is currently in the data collection phase. As such, preliminary results are not yet available. Instead, the poster will present the theoretical framework for the proposed single-switch enhancement, citing related works in literature. We hypothesize that likelihood of switch use during scanning can be estimated from physiological signals with noise filtering from the physical signals, due to the differential treatment of target versus non-target items in scanning by the switch user.
Conclusion: Real-time processing of physiological and physical signals may be able to enhance the reliability of single-switch scanning.