Implemented hardware and software optimizations to enhance the transformer algorithm performance on NeuroSoC, ensuring high efficiency and rapid processing capabilities
This work presents a prototype of a neuromorphic tactile sensing platform configured as a glove demonstrator. The core of the entire system are the tactile sensing and processing chips, which utilize event-driven approaches, thus, mimicking the power-efficient sensing and processing seen in the human somatosensory system and the brain.
In this study, we enriched the methods of physiological-signal- based pain classification by introducing deep Recurrent Neural Network (RNN) based hybrid classifier.
We focus on the node selection problem of grounded Laplacian matrix