BreathTrack: Detecting Regular Breathing Phases from Unannotated Acoustic Data Captured by a ...
summary
A smartphone microphone model learns inhalation and exhalation by transferring labels from IMU signals using a teacher-student transfer where the teacher is a signal processing pipeline and the student is a deep neural network, removing manual audio annotation. Trained on data from 131 subjects, it detects phases with 77.33% accuracy, estimates fractional inspiratory time, I:E ratio and breathing rate with ~92.1%, 86.8% and 91.7% accuracy respectively, and separates patients from healthy users up to 76%.
AI-written summary · about
related talks
42mMachine learning with ventilator data
20mExploring the hidden potential of sound data
01h 14mExploring the hidden potential of sound data
01h 01mVocal biomarkers for disease detection
01h 13mHealth Monitoring with Wireless Sensors and Machine Learning
22mUbicoustics: Plug-and-Play Acoustic Activity Recognition
06mBodyCompass: Monitoring Sleep Posture with Wireless Signals
45mTeaching Machines How to Smell Disease — Lessons Learned