Biometric authentication technology has become increasingly common in our daily lives as information protection and control regulation requirements have grown worldwide. A biometric system must be simple, flexible, efficient, and secure from unauthorized access. The most suitable and flexible biometric traits are the face, fingerprint, palm print, voice, electrocardiogram (ECG), and iris. ECGs are difficult to falsify among these biometric traits and are less attack-prone. However, designing biometric systems based on ECG is very challenging. The major limitations of the existing techniques are that they require a large amount of training data and that they are trained and tested on an on-person database. To cope with these issues, this work proposes a novel biometric authentication scheme based on ECG detection called BAED. The system was developed based on deep learning algorithms, including a convolutional neural network (CNN) and a long-term memory (LSTM) network with a customized activation function. The authors evaluated the proposed model with on-and off-person databases including ECG-ID, Physikalisch-Technische Bundesanstalt (PTB), Check Your Bio-signals Here Initiative (CYBHi), and the University of Toronto Database (UofTDB). In addition to the standard performance parameters, certain key supportive identification parameters such as FMR, FNMR, FAR, and FRR were computed and compared to increase the model’s credibility.The proposed BAED system outperforms prior state-of-the-art approaches.