Classifying Heartbeat Arrhythmia using novel features (AR coefficients + RR distance).
Full explanation available in this notebook (You can explore it on your browser!).
In this project, I implement a heartbeat arrhythmia classification algorithm to separate normal heartbeats (N) from Ventricular Ectopic Beats (VEB). This work is based on the paper "Heartbeats classification using QRS and T waves autoregressive features and RR interval features", Adnane M, Belouchrani A. (see references).
The features used to train the model are:
- Auto-regressive coefficients of the QRS complex and T wave of the electrocardiogram.
- RR interbeat distance.
Normal Beats: 
VEB Beats: 
- A 3rd order high-pass Butterworth filter is used to eliminate DC componant and baseline wander.
- A 3rd order band-reject Butterworth filter is used to eliminate the 60 Hz AC interference.
- A 4th order low-pass Butterworth filter is used to high-freq artifacts like EMG noise.
- We use the Levinson algorithm to extract AR coefficients from ECG signal.
- A basic substraction gives the RR interdistance.
SVM classifiers give good results for now. I'll stick to them while trying to improve hyper-parameters before going to anything else (e.g. Neural Nets).
For the time being (2017), results are reaching those obtained by state-of-the-art methods with accuracy metric 97.02% (overall), and 98.86% for subject specific scheme (See references).