Hardware implementation of AI-based fuzzy logic for real-time sign language interpreter robotic control using FPGA

Communication barriers between hearing-impaired individuals and non-sign language users remain a significant challenge in modern society. Existing sign language recognition systems primarily focus on unidirectional communication and are often limited by high latency, excessive power consumption, and lack of robustness in uncertain environments. This paper proposes a novel FPGA-based Bidirectional Robotic Sign Language Communication System (BRSLCS) integrating Artificial Intelligence (AI), fuzzy logic, and robotic control for real-time assistive communication. The proposed architecture supports both forward communication, where sign language gestures are translated into text or speech, and reverse communication, where text or speech is converted into robotic sign language gestures through an actuated robotic hand. In the forward communication flow, gesture images captured using a camera are processed through preprocessing, feature extraction, AI-based classification, and fuzzy inference stages. In the reverse communication flow, speech or textual inputs are interpreted using Natural Language Processing (NLP) techniques and mapped into robotic sign gestures controlled using FPGA-based actuation logic. The integration of fuzzy logic improves classification robustness and interpretability under uncertain conditions such as illumination variation, gesture ambiguity, and motion blur. FPGA implementation provides deterministic timing, low latency, hardware parallelism, and energy efficiency suitable for embedded real-time systems. Experimental evaluation demonstrates that the proposed system achieves 92-95% recognition accuracy with latency between 10-25 ms while consuming only 4-6 W power, outperforming conventional CPU and GPU-based implementations. The proposed BRSLCS provides a scalable and practical framework for inclusive communication systems in healthcare, education, public services, and human-robot interaction environments.