Embedded System for Bimodal Biometrics with Fiducial Feature Extraction on ECG and PPG Signals

This paper presents an embedded system implementation of bimodal biometrics at features level using electrocardiographic (ECG) and photoplethysmographic (PPG) biosignals with a fiducial feature extraction based on statistical parameters. One-way ANOVA test was applied before the authentication process in order to determine the most relevant features within the ECG/PPG obtained data. L1 distance classification together with the optimized feature extraction provided a hardware-economic, yet efficient, method to be implemented in an embedded system. The biometric system was fully implemented in a Zynq-7000 SoC FPGA, obtaining a processing time of 1.49 msec on feature generation and classification stages. An Equal Error Rate EER=0.066 in average was obtained in the bimodal case, confirming through ROC curves the advantage of using the fusion of both signals versus each one used separately.