Chronic stress can lead to significant health complications, including anxiety, depression, hypertension, cardiovascular diseases, sleep disturbances, and headaches. In this study, we propose a robust hybrid deep learning model that combines Convolutional Neural Networks (CNN) with Bidirectional Long Short-Term Memory (BiLSTM) layers to detect mental stress using physiological signals such as ECG, EDA, three-axis acceleration, heart rate variability, and respiration (RESP).The CNN layers are designed to extract intrinsic features from the input signals, while the BiLSTM layers capture temporal dependencies within the time-series data. We evaluate the model's performance on three publicly available datasets: WESAD, SWELL-KW, and Stress-Lysis. The model demonstrates high classification accuracy across multiple scenarios, including 2, 3, 4, and 8-class stress level detection tasks. To further validate our approach, we conduct an ablation study, comparing the full hybrid model against a variant without BiLSTM layers. The results indicate a clear performance gain with the inclusion of BiLSTM, highlighting the model’s ability to generalize and effectively learn complex relationships in the data. This research emphasizes the potential of deep learning techniques for mental stress detection, providing valuable insights for the development of preventive healthcare solutions.