Enhancing Gesture Recognition: Exploring CNN-LSTM Networks for Dynamic Dual- Hand Gestures

Various hand, lip, and facial gestures are often used by millions of people who have speech and hearing difficulties to communicate. Indian Sign Language (ISL) has a highly extensive vocabulary that differs from region to region. There is no publicly available dataset on serial gestures for ISL. The dynamic hand gesture dataset, which the authors have made public, is broken down into 33 categories, including days of the week, months of the year, and gestures used frequently. This dataset was compiled using a burst shot data gathering technique. A smaller portion of the dataset is used to enable fast examination, with 12 classes representing the 12 months of the year. It is somewhat intricate due to its recommendations, and each category often contains 5–6 movements. The method for installing the particular neural network on hardware is also covered in this publication. The suggested concept is designed to run on hardware that consumes less power. The proposed model is further compared with other sequential architectures to find the best model for hand gesture motion detection.