ECG based authentication using Autocorrelation and Artificial Neural Networks

Biometrics refers to metrics or parameters related to human characteristics. Several Biometric security systems and authentication proceduresare already in use ranging from fingerprint scanners to facial recognition software. Even with the rapid advancement of these security systems, they are still susceptible to frauds. The proposed ECG based authentication (ECGA) could prove to be highly secure and cost effective, if employed correctly. This is a non-invasive technique. The Electrocardiogram (ECG) obtained is unique to different individuals i.e., no two persons have same cardiac rhythm. ECGA aims to exploit this fact to provide unique security identity to all the individuals which is used for authentication process. A new authentication strategy is being employed here, where Time domain ECG signal processing is performed involving filtering, peak detection, ECG waveform segmentation, autocorrelation and amplitude normalization. Further the processed signal is fed to the Neural Networks, powers of Neural Networks are harnessed to create a supervised system that can identify individuals by matching the real time sample to the template present in the template database. This project results on a small dataset and proves that ECG appears to be a viable trait for real-world biometric scenarios.