An embedded FPGA accelerator for a stand-alone dual-mode assistive device

This paper presents a stand-alone Dual-mode Tongue Drive System (sdTDS) which is designed for people with severe disabilities to control their environment using their tongue motion and speech. The sdTDS detects user's tongue motion using a magnetic tracer placed on tongue and an array of magnetic sensors embedded in a wireless headset and at the same time it can capture the user's voice using a small microphone embedded in the same headset. A real-time FPGA-based local processor is proposed which can perform all required signal processing to convert raw data generated by magnetic sensors and microphone to user commands at sensor side, rather than sending all raw data out to a PC or smartphone. The proposed sdTDS significantly reduces the transmitter power consumption and subsequently increases the battery life. Assuming the sdTDS user issues one command every 20 ms, implementing the proposed local processor reduces the data volume that needs to be wirelessly transmitted from 25.6 kb/s to 0.3 kb/s. To evaluate the functionality and performance of the sdTDS processor, it has been implemented on a Xilinx Zynq SoC device (ARM+Artix FPGA) and at frequency of 100 MHz, it consumes 2 mJ energy. The detection accuracy is 96.6% for tongue motion, and 97.5% for speech recognition.