Utilizing deep neural nets for an embedded ECG-based biometric authentication system

This work presents a low-power, embedded ECG pattern recognition system for the purpose of biometric authentication. We believe that ECG coupled with a secondary biometric marker such as fingerprint will play a key role in wearable security as wearables' popularity continues to grow. The key objective of this work is to implement a system that is reliable, robust, and fast while maintaining a low area and power footprint. A streamlined approach was devised that utilized neural networks to both identify QRS complex segments of the ECG signal and then perform user authentication on these segments. When tested on 90 individuals, the system is able to achieve 99.54% accuracy for QRS complex identification, and, on average, 99.85% sensitivity, 99.96% specificity, and 0.0582% EER for user identification. When implemented on an Artix-7 FPGA, the entire design occupies 1,712 slices (5%) and 978.7 KB of memory and dissipates 31.75 mW of total chip dynamic power when running at 12.5 MHz.