Gesture Recognition for Human Contactless Computer Interaction

This research introduces a novel deep-learning model that can recognize hand signals in real-time, making contactless human-computer interaction possible. With the aid of MediaPipe, users can easily track their hands while using a hybrid deep learning model that incorporates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks for temporal sequence modeling. Optimized preprocessing through segmentation and feature normalization is achieved through the use of real-time hand detection by MediaPipe, leading to highly robust recognition across a wide range of environments. Both static and dynamic gestures can be classified accurately due to the LSTM module's temporal dependencies and spatial features extracted by the CNN component. Extensive testing has established that the model is highly accurate, scalable, and low-latency. To overcome these challenges, this paper suggests a hybrid CNN-LSTM-based model with the MediaPipe for effective hand tracking and gesture segmentation. The objective is to provide low-latency, real-time detection of static and dynamic gestures for smooth control of YouTube interfaces. This suggested approach is intended to improve spatial-temporal modeling with the provision of lightweight deployment that is suitable for real-time applications.