HeartQuake: Accurate Low-Cost Non-Invasive ECG Monitoring Using Bed-Mounted Geophones
summary
A bed-mounted system uses an SM-24 geophone plus amplifier and Raspberry Pi to capture mattress-transmitted heartbeat vibrations, applies a 5–30 Hz band-pass to remove sensor noise, and feeds signals to a stacked Bi-LSTM to reconstruct full ECG waveforms. Trained on PhysioNet (20 subjects) then fine-tuned on 21 subjects, it achieves 13 ms average error and 99% peak detection, estimates RR and QRS timings within a few milliseconds, degrades with motion or external vibration but improves with personalization, and clinical evaluation with 11 doctors finds HeartQuake signals indistinguishable from clinical ECG samples.
AI-written summary · about
related talks
07mCardiacWave: A mmWave-based Scheme of Non-Contact and High-Definition Heart Activity Computing
01h 13mHealth Monitoring with Wireless Sensors and Machine Learning
46mAnalyzing the ElectroCardioGram (ECG)
05mRF Vital Sign Sensing Under Free Body Movement
21mViBand: High-Fidelity Bio-Acoustic Sensing Using Commodity Smartwatch Accelerometers
28mHow Smart is the ECG? Lessons from Advanced ECG Analytics
06mWR-Hand: Wearable Armband can Track User's Hand
05mMy heart depends on your code