EchoBreath: Continuous Respiratory Behavior Recognition in the Wild via Acoustic Sensing on Smart Glasses

要旨

Monitoring the occurrence count of abnormal respiratory symptoms helps provide critical support for respiratory health. While this is necessary, there is still a lack of an unobtrusive and reliable way that can be effectively used in real-world settings. In this paper, we present EchoBreath, a passive and active acoustic combined sensing system for abnormal respiratory symptoms monitoring. EchoBreath novelly uses the speaker and microphone under the frame of the glasses to emit ultrasonic waves and capture both passive sounds and echo profiles, which can effectively distinguish between subject-aware behaviors and background noise. Furthermore, A lightweight neural network with the 'Null' class and open-set filtering mechanisms substantially improves real-world applicability by eliminating unrelated activity. Our experiments, involving 25 participants, demonstrate that EchoBreath can recognize 6 typical respiratory symptoms in a laboratory setting with an accuracy of 93.1%. Additionally, an in-the-semi-wild study with 10 participants further validates that EchoBreath can continuously monitor respiratory abnormalities under real-world conditions. We believe that EchoBreath can serve as an unobtrusive and reliable way to monitor abnormal respiratory symptoms.

受賞
Honorable Mention
著者
Kaiyi Guo
shanghai jiao tong university, shanghai, China
Qian Zhang
Shanghai Jiao Tong University, Shanghai, China
Dong Wang
Shanghai Jiao Tong University, Shanghai, China
DOI

10.1145/3706598.3714171

論文URL

https://dl.acm.org/doi/10.1145/3706598.3714171

動画

会議: CHI 2025

The ACM CHI Conference on Human Factors in Computing Systems (https://chi2025.acm.org/)

セッション: Earable and Hearable

Annex Hall F206
5 件の発表
2025-04-28 20:10:00
2025-04-28 21:40:00
日本語まとめ
読み込み中…