Stroke is a leading cause of death and disability worldwide. Early detection of stroke symptoms, such as facial weakness, is critical for timely treatment. The FAST method (Facial drooping, Arm weakness, Speech difficulty, and Time) is widely used for identifying these symptoms. This study introduces a novel system for detecting facial weakness in stroke patients using Feedback Cascade Regression for facial landmark alignment and Support Vector Machine (SVM) for classification. The proposed method calculates 29 facial tilt features to assess facial symmetry and employs Principal Component Analysis (PCA) for dimensionality reduction. The SVM model, optimized with a grid search for parameters (C = 100, gamma = 0.1), achieved a 95% accuracy in distinguishing between stroke-affected and normal faces. The findings suggest that integrating advanced pre-processing techniques with SVM classification can significantly enhance facial weakness detection. This research contributes to improving the accuracy and robustness of stroke detection systems. Implications: The proposed system can be integrated into clinical practice for early stroke detection, potentially improving patient outcomes through timely intervention.