Low-Power Mobile Sign Language Recognition: Real-Time Optimization for Resource-Constrained Devices

Abstract

In recent years, the demand for effective communication tools for the hearing-impaired community has surged, prompting advancements in sign language recognition technologies. This paper presents a novel approach to low-power mobile sign language recognition, focusing on real-time optimization for resource-constrained devices. We propose a lightweight network architecture designed to operate efficiently on mobile platforms, such as ARM-based devices, while maintaining high accuracy in gesture recognition. Our method leverages low-cost sensors and digital signal processing techniques to capture and interpret sign language gestures in real-time. By employing a combination of handcrafted descriptors and deep learning algorithms, we enhance the model's ability to recognize a diverse range of signs with minimal computational overhead. Extensive experiments demonstrate that our system achieves competitive performance compared to state-of-the-art models, with a significant reduction in power consumption and latency. Furthermore, we explore the deployment of our recognition system on mobile devices, ensuring seamless integration into everyday applications. The results indicate that our approach not only facilitates effective communication for the hearing-impaired but also promotes accessibility and inclusivity in various environments. This research contributes to the ongoing efforts to bridge the communication gap between the hearing and hearing-impaired communities, paving the way for future developments in mobile sign language recognition technologies.