Explorations on Breathing Based Text Input for Mobile Devices
Jackson F. Filho, Thiago Valle, Wilson Prata · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS) · doi:10.1145/2661334.2661425
Summary
This paper presents progress on PuffText, a breathing-based text input system for mobile phones designed for people with motor disabilities who cannot handle a mobile device manually. The system uses the phone's built-in microphone to detect "puffs" of breath, which serve as the sole input mechanism for navigating and selecting characters from a dynamically generated spinning keyboard. The original PuffText system detected three distinct puff events: a single short puff (stops the spinning keyboard and zooms the character under the cursor), a double short puff (selects the character and adds it to the text), and a long puff (switches between letter, number, and punctuation keyboards). This paper reports on improvements to the system based on usability testing conducted over 6 months in 8 rounds with at least 6 participants per round (3 disabled, 3 able-bodied) in a usability lab with 7 cameras and 9 microphones. Based on user feedback — some participants had difficulty understanding the spinning "dial" metaphor and found puff detection unreliable — the researchers redesigned the interface to use a linear character display rather than a spinning dial, and improved the puff detection algorithm by applying a bandpass filter (focusing on 100-500 Hz range) and using signal energy measurement with dynamic thresholds to better distinguish intentional puffs from ambient noise and normal breathing. The system requires no additional hardware beyond the phone itself, making it a low-cost, always-available input method.
Key findings
The redesigned PuffText system improved upon the original in two key ways. First, the interface was changed from a spinning dial metaphor to a linear display, addressing user confusion reported in earlier testing. Second, the puff detection algorithm was enhanced with a bandpass filter isolating the 100-500 Hz frequency range characteristic of puff sounds, combined with signal energy analysis using dynamic thresholds — making detection more reliable and less susceptible to environmental noise. The usability testing over 6 months with mixed disability and able-bodied participants validated that the improvements made the system more efficient and user-friendly compared to the original implementation. The system positions breathing-based input as a valuable hands-free and silent alternative to speech recognition for text entry on mobile phones, noting that speech recognition — while widely available — may not be suitable in all situations (noisy environments, privacy concerns, or for users with speech impairments alongside motor disabilities). The approach of using the phone's existing microphone eliminates the need for specialised sip-and-puff hardware, which is typically expensive and requires tube-based physical contact.
Relevance
This work explores an underserved area of accessible text entry: input methods for people who cannot use their hands at all and may also be unable to use speech recognition. Traditional sip-and-puff controllers require dedicated hardware and physical contact with a tube, making them less portable and hygienic than a microphone-based approach. By using the phone's built-in microphone, PuffText requires no additional equipment — the user simply blows toward the phone. For accessibility practitioners, this represents a genuinely minimal-hardware approach to hands-free mobile input. The iterative design process involving disabled participants over multiple rounds demonstrates good practice in assistive technology development. However, the fundamental speed limitation of a scanning-based single-input system (one character at a time via puffs) means this approach is best suited for short communications rather than extended text composition. The work is also relevant to the broader question of how many distinct input signals can be reliably extracted from breathing — a question relevant to brain-computer interface research and to designing input for people with the most severe motor impairments who retain only breath control.
Tags: text entry · alternative input · motor disability · sip-and-puff · mobile accessibility · breathing input · assistive technology