Conception and Experimentation of a Communication Device with Adaptive Scanning
Souhir Ghedira, Pierre Pino, Guy Bourhis · 2009 · ACM Transactions on Accessible Computing (TACCESS) · doi:10.1145/1497302.1497304
Summary
This paper presents an adaptive algorithm for optimizing scanning delay (Tscan) in switch-operated AAC devices, addressing one of the fundamental challenges in scanning-based communication: finding the right balance between speed and error rate. For users with severe motor disabilities and speech disorders, single-switch scanning systems are often the only viable communication method, but fixed scanning speeds inevitably compromise either efficiency (too slow) or accuracy (too fast). The researchers developed their approach using the EDiTH system (Digital Teleaction Environment for People with Disabilities), a comprehensive AAC platform created at the University of Lorraine that provides communication, reading, email, and environment control via column-row scanning. Central to their methodology is an adaptation of Card et al.'s Model Human Processor (MHP), which decomposes reaction time into perceptual (~100ms), cognitive (~70ms), and motor (~70ms) components. Through extensive log file analysis, they identified a three-zone behavioral model: an "advance zone" (response times under 100ms) indicating anticipations or errors, a "central zone" (100-400ms) representing normal MHP behavior, and a "lead zone" (over 400ms) suggesting fatigue or concentration difficulties.
Key findings
The study validated that motor response times for people with disabilities range from 100-300ms, compared to 60-80ms for able-bodied users, while perceptual and cognitive processing times remain similar across groups. This finding is significant because it means adaptive systems should focus on accommodating motor variability rather than assuming overall cognitive differences. The adaptive algorithm uses the frequency of responses in the advance zone (N100) as its primary indicator: when too many responses fall below 100ms, the scanning delay is increased; when few do, it can be decreased. Testing with participants including people with tetraplegia, cerebral palsy, locked-in syndrome, ALS, and cranial trauma demonstrated that the algorithm stabilizes Tscan around 180-250ms for expert users—significantly faster than typical clinically-set values of 600-1200ms. Critically, the algorithm works purely on timing data and requires no knowledge of the scanning device itself, making it transferable across different AAC systems. However, approximately 30% of participants (those with cerebral palsy, locked-in syndrome, or cranial trauma) did not conform to the three-zone model due to motor control or attention difficulties.
Relevance
This research has direct implications for AAC device configuration and clinical practice. The finding that optimal scanning delays can be much faster than typically prescribed—and that they should vary dynamically with user fatigue and attention—challenges the common practice of setting a fixed delay at initial assessment. Clinicians should consider adaptive scanning as a standard feature rather than an advanced option. The device-independent nature of the algorithm means it could potentially be implemented as a software layer on top of existing AAC systems. For developers, the three-zone behavioral model provides a validated framework for detecting when scanning speed is mismatched to user capabilities. The research also highlights that one-size-fits-all approaches fail for approximately one-third of users, emphasizing the need for multiple adaptation strategies based on disability type and motor control characteristics.
Tags: AAC · scanning · switch access · motor disabilities · adaptive systems · human-computer interaction · ALS