Human identification by simultaneous recording of acceleration and ECG data

Abstract

This paper presents a method of human identification based on ensemble empirical mode decomposition (EEMD) of an one-lead electrocardiogram (ECG) signal and by box approximation geometry of reconstructed attractors in latent space of a signal measured by an accelerometer located on the waist. Preprocessing of the ECG signal eliminates effects of noise and heart rate variability. The ECG signal is decomposed into a number of intrinsic mode functions (IMFs) and significant heartbeat signal features are extracted using Welch spectral analysis. Human gait is considered a dynamical system and the features are the eigenvalues of the reconstructed attractor in the odd principal dimensions obtained using the Singular Spectrum Analysis methodology. The Knearest neighbours (K-NN) method is applied as the classifier tool.