LyreFlute: Sensor-Based Musical Experience for COPD Rehabilitation
Andreia Valente, Claudio Cotto, Mark Billinghurst, Kunal Gupta · 2026 · Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA '26) · doi:10.1145/3772363.3799123
Summary
LyreFlute is a sensor-equipped musical instrument paired with a 3D rhythm game that reframes at-home COPD (Chronic Obstructive Pulmonary Disease) breathing exercises as a culturally grounded, narrative-driven experience. The authors motivate the work with a stark adherence problem: at-home COPD breathing programs show only 16-32% adherence, with patients citing monotony, lack of feedback, and difficulty knowing whether they are performing techniques correctly. LyreFlute responds with three coordinated moves. First, a custom 3D-printed flute (322 mm long, 32 mm diameter) embeds two FS3000 airflow sensors - one inside the flute for oral exhalation and one mounted on a ramp angled toward the nose for nasal inhalation - physically enforcing the nose-in / mouth-out pattern central to pursed-lip breathing. Four copper finger rings on an MPR121 capacitive touch sensor map to a one-octave chromatic scale via a 4-bit binary encoding. A XIAO ESP32S3 with a BNO08X IMU streams data over Bluetooth to a Unity game. Second, a three-day rehabilitation arc (a condensed analogue of the typical multi-week clinical programme) progresses through sustained pursed-lip phrases on Day 1, interval-style breath groups on Day 2, and denser passages with shorter recovery on Day 3, mirroring established interval-training principles. Third, the experience is wrapped in a story about restoring the Australian superb lyrebird's habitat after the 2019-2020 bushfires - correct breathing extinguishes flames, clears smoke, and regrows vegetation. Audio is generated by routing flute sensor data through a custom .NET VST host running a Suling Sunda flute preset.
Key findings
This is a CHI Interactivity demo paper rather than a clinical study, so the contributions are design and engineering. The authors show that a single instrument can simultaneously serve as a respiratory training tool, a music-making device, and a sensor platform: separating nasal inhalation and oral exhalation into two physically distinct sensors mechanically prevents the most common error in pursed-lip breathing without relying on user discipline. The flute provides about three hours of operation on a 380 mAh lithium-polymer battery and transmits to a desktop running Unity, where breath timing, duration, and intensity are scored in real time and fed back as environmental change in the 3D scene. The three-day musical structure encodes recognisable clinical principles - sustained breath cycles, then alternating interval patterns with recovery breaths, then higher-density patterns with shorter recovery - in a form that reads as gameplay rather than therapy. The authors acknowledge the work has not yet been clinically validated; planned next steps include a 15-day COPD cohort trial with breathing straps and ECG, exhalation-velocity-controlled octave shifts, an Android port to lower cost, and culturally adapted instruments such as the koauau and bamboo flute for different patient populations.
Relevance
For accessibility and digital-health practitioners, LyreFlute is a compact example of how to attack a well-documented adherence problem by changing the form of the activity rather than nagging users into it. The hardware design pattern - using physical sensor placement to enforce a clinically correct movement pattern - is transferable to other rehabilitation contexts where users cannot easily self-monitor technique, including stroke recovery, voice therapy, and post-surgical breathing. The cultural-grounding argument is also worth taking seriously: the authors explicitly plan to swap the instrument and narrative for other patient cohorts rather than ship a single 'universal' design, which is a useful counter to one-size-fits-all health technology. Limitations to keep in mind: there is no efficacy evidence yet, the demo is a three-day compression of a multi-week protocol, the device assumes sufficient fine-motor control to cover four finger holes accurately, and the shared mouthpiece raises hygiene and accessibility concerns for clinic deployment that the demo addresses with disposable mouthpieces but a home product would need to revisit.
Tags: COPD · pulmonary rehabilitation · breathing exercises · biofeedback · serious game · gamification · tangible interaction · health informatics · adherence · embodied interaction