SVM-based Emotion Recognition for Detecting Early Signs of Psychological Distress

Timely identification of psychological discomfort is crucial for averting serious mental health disorders, including anxiety and depression. This paper presents a Support Vector Machine (SVM)-based emotion identification system that utilizes multimodal data, including facial expressions, voice attributes, and physiological signals, to detect early indicators of distress. The system was trained on a dataset including 5,000 labeled samples, achieving an accuracy of 96.8%, a precision of 95.9%, and a recall of 95.2% in the classification of distress-related emotions. Feature extraction was accomplished using Mel-Frequency Cepstral Coefficients (MFCCs) for speech, Histogram of Oriented Gradients (HOG) for face expressions, and Heart Rate Variability (HRV) for physiological data. The proposed model improves traditional classifiers’ accuracy. Real-time validation on an independent test set of 1,000 samples verified a detection rate of 94.5%, illustrating the system's resilience and dependability. The findings demonstrate that the proposed SVM-based method is very efficient in detecting psychological discomfort and may be included in real-time mental health monitoring systems.