The paper presents a comparative analysis of Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) for selecting optimal geometrical and statistical features in online handwriting recognition. It emphasizes the challenges of feature extraction in achieving high recognition accuracy due to variations in writing styles and proposes a hybrid extraction technique that integrates multiple characteristics of handwritten characters. The results indicate that PSO outperforms GA in terms of feature selection efficiency, training time, and recognition accuracy.