Making Speech-Based Assistive Technology Work for a Real User

Abstract

We present a customized speech-activated email system that is the product of efforts focused on a single target user with high speech recognition error rates. The system, which includes offthe-shelf and custom hardware and software, allows the user to use speech to send emails with recorded audio attachments. Over the past 16 months, our target user has sent and received hundreds of emails and has integrated the system into his daily life. Key factors contributing to the long-term adoption of the device include our extended efforts to understand the target user over multiple years, iterative design, and the collaboration of our multidisciplinary team of assistive technology (AT) designers, clinicians, software developers, and researchers. Overall, we ask: if we set our sights on developing and supporting a technology that someone will actually use daily, what can we learn? We share our approach, system design, user observation and findings, with implications for speechbased AT research and development.