Paper Info
Title
Separating heartbeats from multiple people on one bed using geophones: PhD forum abstract.
Abstract
Sensing bed vibrations caused by heartbeats has shown great potentials in detecting and monitoring a person's heartbeats during sleep, without requiring special mattress or sheets, or assuming certain sleeping position/posture. Earlier work has studied how to use this method to detect heartbeats when a single subject is on the bed, and in this study, we aim to separate the heartbeats when multiple subjects share the same bed and the vibration signals are mixed together. Our heartbeat separation algorithm is based upon signal unmixing via time-frequency masking [4], which was originally designed to extract individual voices from two audio mixtures. Though these two problems have similarity, separating heartbeat signals is much harder and poses new challenges, mainly because heartbeat signals have a much smaller frequency range than audio signals, fluctuate considerably from beat to beat, and propagate through a mattress that has much more complex propagation properties than the air. In this study, we address these challenges by carefully designing the signal processing algorithms, especially in phase correction, filtering, window size choice, etc. Through detailed experimentation, we show that our technique can accurately separate two heartbeats (the most common case) using two vibration sensors (geophones in our case) - with an average estimation error below 2 beats per minute.
Year
DOI
Venue
2017
10.1145/3055031.3055051
IPSN
Keywords
Field
DocType
Heartbeat Sensing, Sleep Monitoring, Signal Processing, Blind Source Separation
Audio signal,Signal processing,Masking (art),Computer science,Real-time computing,Beat (music),Artificial intelligence,Blind signal separation,Computer vision,Heartbeat,Geophone,Filter (signal processing),Speech recognition
Conference
Citations 
PageRank 
References 
0
0.34
1
Authors
1
Name
Order
Citations
PageRank
Zhenhua Jia1305.33