Heart Beat Detection Project using Emfit Matress sensor
Overview
Developing reliable methods to detect heartbeats using non-invasive EMFiT mattress sensors.
Background
Conventional heart monitoring can be intrusive, motivating alternative solutions embedded in everyday sleeping environments, mostly for screening purposes of heart disease, and sleep measurement (sleep architecture, quality, and stress / recovery analysis).
Conventional BCG is studied in time. Although there are methods to detect heartbeats, also other domains, the method developed produced a PPG-like signal that is independent on the position of the patient.
Aim
To accurately identify heartbeats from mattress sensor signals, ensuring comfort and minimal disruption. Explore alternative methods and exploit information in the spectral domain.
Methods
Applying spectral analysis and NMF-based source separation (non-negative tensor factorization of the amplitude spectrogram) techniques to isolate and demodulate heart pulse signals.
Results
Preliminary tests show promising accuracy in heartbeat detection, independent of patient position or movement.
Resources
The project produced the following publications
-
Heart pulse demodulation from Emfit mattress sensor using spectral and source separation
techniques.
. JM Perez-Macias, M Tenhunen, A Värri, SL Himanen, J Viik. 2022 Computing in Cardiology (CinC)