Recognition of Indian sign language alphanumeric gestures based on global features
B. V. Poornima, S Srinath, S. Rashmi, R Rakshitha · 2023
This manuscript presents a novel approach for accurately recognizing Indian sign language alphabets and numbers (1-9), which play a vital role in communication for individuals with hearing impairments in India. We conducted extensive experiments on our self-created dataset named "ISL 2022: A novel dataset creation on Indian sign language". This dataset contains images with non-uniform and complex backgrounds, reflecting the real-world scenarios where sign language gestures are performed. After identifying the region of interest i.e. the hand region from the images, we extract discriminative features using global image descriptors and Hu moments. Global image descriptors capture the global spatial layout of the gesture, providing valuable information about its overall structure and appearance. On the other hand, Hu moments capture shape characteristics essential for differentiating between different gestures. We employ the Random Forest and Naive Bayes algorithms to classify the gestures. The experimental results demonstrate the effectiveness and recognition rate of 98%. By combining the features of both the feature extractors, our approach showcases promising advancements in recognizing ISL gestures, especially considering the challenges posed by diverse backgrounds. The proposed methodology has the potential to contribute significantly to the development of practical and efficient sign language recognition systems.