This study explores the potential of photoplethysmography (PPG) signals as a reliable biometric identification method, emphasizing its significance in addressing security issues through a noninvasive and difficult-to-forge approach. We designed, implemented, and evaluated various models to classify PPG signals, focusing on reliability and accuracy. Using a robust protocol for PPG signal acquisition, signals from 40 individuals were collected. The research explores multiple preprocessing techniques, including data augmentation and normalization to create six distinct datasets. Five deep learning models, including 1D Convolutional Neural Network (1D CNN), Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), Gated Recurrent Unit (GRU), and Dense Neural Network (DNN), are implemented and compared for their classification performance. Among these, the 1D CNN model demonstrated superior classification accuracy, achieving 99.94%. This work reinforces the efficacy of PPG signals in biometric recognition and highlights the potential of advanced deep learning techniques in enhancing recognition accuracy.