Real-Time American Sign Language Recognition: A Comparative Study of Machine Learning and Deep Learning Approaches

American Sign Language (ASL) Recognition systems function as a bridge between the deaf population and hearing population. This paper presents a comparative analysis of Machine Learning (ML) and Deep Learning (DL) approaches for ASL recognition in real-time settings. The ML based approach involves keypoint feature extraction and classification using Multi layer Perceptron (MLP), XGBoost, and Support Vector Machine (SVM). In contrast, the DL based approach employs pre-trained convolutional neural networks of EfficientNet B0, VGG16, and ResNet50 on segmented hand images, classifying both alphabets (A-Z) and numerals ($0-9$). Both approaches utilize systems with a webcam for real-time ASL recognition, with comparative analysis based on accuracy. The results show that ML can be a lowresource solution for ASL recognition, while DL offers higher accuracy at the cost of increased computational demands. In the ML approach, Support Vector Machine (SVM) achieved the highest accuracy of 95.0 %, whereas in the DL approach, ResNet50 attained the best performance with an accuracy of 98.3 %.