Reconfigurable SoC Platform for Edge AI in Assistive Technologies

This book examines how current technology can facilitate communication for Deaf/Hard of Hearing individuals. The book outlines a system for recognizing Indian Sign Language (ISL) in real-time through the use of a Xilinx Zynq-7000 chip that contains both an ARM processor and an FPGA. The development of a small 8-bit CNN model to use within the FPGA allows for quick recognition of sign language gestures offline.By eliminating the need for internet-based processing, and therefore any associated delays, such a system would make it feasible to use in many situations with portable devices. Testing indicates an accuracy rate of 99.54% and minimal power consumption while utilizing only a fraction of the potential programmable space within the FPGA. This demonstrates that developing assistive technologies with the proper combination of hardware and software results in devices capable of delivering high levels of performance with low energy consumption.