This paper introduces an innovative approach to recognizing stress and meditation emotions, building upon prior research in the field. It focuses on Photoplethysmogram (PPG) signals, known for their sensitivity to stress-related physiological changes, and employs Continuous Wavelet Transform (CWT) analysis for a comprehensive exploration of stress-related patterns. While previous works in stress detection have primarily concentrated on various physiological signals and algorithms, this study stands out by emphasizing PPG signals, which provide unique insights into cardiovascular stress responses. The research combines CWT, Convolutional Neural Networks (CNNs), and Kalman filtering to enhance stress analysis, representing a notable advancement in wearable technology. An application of this research is in developing alarm systems for mental stress. The proposed method efficiently processes PPG signals by employing a Kalman filter for noise reduction, segmenting data into windows, and utilizing CWT-based CNNs for feature extraction and classification. Using the WESAD dataset, encompassing diverse stress scenarios, the study demonstrates the effectiveness of this approach. The CNN classifier achieves an impressive test accuracy of approximately 95.60%, highlighting its suitability for real-time stress detection applications. Its consistent performance over multiple epochs suggests practical implementation potential. Additionally, the presented confusion matrix and Receiver Operating Characteristic (ROC) curve validate the model’s discriminative power, further underscoring its potential for stress classification.