Deep Learning Techniques Based Secured Biometric Authentication and Classification using ECG Signal

The use of biometric authentication in our daily lives has progressively expanded, owing to the rising global requirements for information security and regulatory oversight. Face, palm print, voice, fingerprint, iris, and Electrocardiogram (ECG) are among the most adaptive and effective biometric features. However, contemporary methodologies frequently necessitate a large amount of training data, which is typically collected from an on-site ECG database for training and testing. To solve these limitations, Deep Learning (DL) techniques such as Recurrent Neural Network (RNN) and Bidirectional Long Short Term Memory (Bi-LSTM) is proposed for secure biometric authentication system. The preprocessing method is utilized for denoising the noises in every individual ECG signals. Then the preprocessing, the data are split into two phases for classification by utilizingRNN and Bi-LSTM. In this proposed method, the databases of on-person and off-person such as ECG-ID, Check Your Bio-signals Here Initiative (CYBHI), Physikalisch-Technische Bunde-sanstalt (PTB), and University of Toronto Database (UofTDB) are used. The proposed RNNBi-LSTM approach accomplished an overall value in on-person datasets from metrics such as 99.81% accuracy, 99.79% sensitivity, 99.86% specificity, and 99.78% positive predictivity. Similarly, in off-person dataset attains 98.76% accuracy, 98.78% sensitivity, 98.84% specificity and 98.73% positive predictivity.