Problem identified

• Increased adoption of wearable heart-rate sensors opens opportunity to understand mobile health behavior
• Classification of arrhythmias and other heart complications saves lives and costs

Solution created

• Pipeline to input, clean, de-trend, process, and fit ballistocardiogram (BCG) and electrocardiogram (ECG) waveforms to Hidden Markov Models.

Technologies used

• MATLAB + DSP Toolbox
• Python hmmlearn

De-trended signals and labelled R and S peaks

Overlaying S-to-S ECG complexes

HMM applied to processed ECG signals