MudraNet Kannada Sign Language Recognition Using Machine Learning
Minavathi Minavathi, Rahul H U, J S GaganaShree, T S Darshan, R Varshith, P Monika · 2024
Sign language is the only trustworthy means of communication for deaf people worldwide.. There are a lot of suggested methods for automatic language recognition in the literature. Nevertheless, the scientific community paid less attention to ARSL (Arabic Sign Language) than to ASL (American Sign Language). In this study, we suggest a new approach that simplifies the hand identification process without requiring the deaf to wear cumbersome gear like gloves. Gesture derived from 1D photos forms the basis of the system. This operation is accomplished by using the Hands module technique in Mediapipe, which extracts features from the hand landmarks. Additionally, the introduced system's accuracy is increased by focusing only on the hand in the whole image using Hands module technique. The pipeline starts by extracting hand landmarks using the Mediapipe Hands model, which provide high-dimensional feature vectors capturing hand spatial configuration. OpenCV assists in image preprocessing and converting raw data for feature extraction. The hand landmarks serve as input features for training a Random Forest classifier, chosen for its effectiveness with high-dimensional data and resilience to overfitting. The classifier is trained on a labeled dataset of hand gesture images. Once trained, it predicts hand gesture classes for new images with reasonable accuracy, evaluated using metrics like classification accuracy and recall.