Hand Sign Recognition using Deep Learning Models

The purpose of this research is to demonstrate a Long Short-Term Memory (LSTM) network-based hand sign recognition system that can recognize temporal correlations in sequential hand gestures. Conventional techniques, such as Convolutional Neural Networks (CNNs), perform poorly when used for dynamic, time-varying motions, but they function well for static image identification. To solve this, we paired a CNN for spatial feature extraction with an LSTM network to handle temporal sequences of hand signs. Each of the X hand gestures that were recorded as a succession of frames over time made up a custom data set of hand sign sequences. The system records user gestures via a real-time input interface, preprocesses the data, and then uses preset output mappings to identify the gesture. With an accurate 90% on the test set, the model proved its capacity to recognize and understand hand signs correctly in real-time applications. Applications involving dynamic hand gestures, such as sign language translation and gesture-based control systems, benefit greatly from the adoption of this LSTM-based technique. Our findings demonstrate that temporal modeling via long short-term memory (LSTM) can greatly enhance gesture recognition systems' performance, particularly in situations when time-based information is essential.