Prizes 2022

Jan. 15, 2023

Misha Mahowald Prizes

The MMP jury has awarded the 2022 Misha Mahowald Prize for Neuromorphic Engineering to a team led by Nitish Thakor at the Johns Hopkins University, for their pioneering work on "Neuromorphic e-Dermis for Restoring Complex Touch" that brings tactile perception to humans with a prosthetic arm.

The MECA jury has awarded the 2022 Mahowald Early Career Award to Chang Gao (TU Delft) for the project "Accelerating Recurrent Neural Networks with Neuromorphic Principles". The project introduces efficient computing using the neuromorphic principles of spatial and temporal sparsity, leading to an energy-efficient accelerator for edge RNN computing. Gao's accomplishments have the potential to revolutionize applications in healthcare, wearables, and intelligent biomedical implants.

See the 2022 prizes press release and read below for details of the 2022 MMP and MECA awards.

2022 Misha Mahowald Prize for Neuromorphic Engineering

Neuromorphic e-Dermis for Restoring Complex Touch

The Neuromorphic e-Dermis Team invented a multilayered, artificial electronic dermis (e-dermis) to restore complex tactile perceptions of pressure and pain to prosthesis users and provide robotic systems with intelligent, neurally-inspired representations of high-dimensional tactile information using neuromorphic representations of touch.

The team members are: Luke Osborn (1) , Andrei Dragomir (2) , Joseph Betthauser (1) , Christopher Hunt (1) , Harrison Nguyen (1) , Rahul Kaliki (3) and Nitish Thakor (1,2) (PI)

1: Johns Hopkins University; 2: Singapore Institute for Neurotechnology, National University of Singapore; 3: Infinite Biomedical Technologies

The work was based at the Neuroengineering and Biomedical Instrumentation Laboratory at Johns Hopkins University.

The e-Dermis work is described in "Prosthesis with neuromorphic multilayered e-dermis perceives touch and pain", Science Robotics 2018 https://www.science.org/doi/10.1126/scirobotics.aat3818

This work has collectively resulted in 34 peer-reviewed publications with a total of 633 citations as of Oct 29, 2022.

Using the multilayered e-dermis, tactile information from a prosthesis grasping an object is transformed into a neuromorphic signal through the prosthesis controller. The neuromorphic signal is used to elicit sensory perceptions of touch and pain to both the prosthesis – enabling intelligent autonomous behaviors – and to the user through transcutaneous stimulation of peripheral nerves – restoring perception and function.
(A) The multilayered e-dermis is made up of conductive andpiezoresistive textiles encased in rubber. A dermal layer of two piezoresistive sensing elements is separated from the epidermal layer, which has one piezoresistive sensing element, with a 1-mm layer of silicone rubber. The e-dermis was fabricated to fit over the fingertips of a prosthetic hand. (B) The natural layering of mechanoreceptors in healthy glabrous skin makes use of both RA and SA receptors to encode the complex properties of touch.
A prosthetic hand with multilayered e-dermis on the fingertips is able to detect touch and pain. Painful sensations trigger an automatic reflex in the prosthesis to release an object.

2022 Mahowald Early Career Award

Accelerating Recurrent Neural Networks with Neuromorphic Principles

During his PhD, Chang Gao developed a series of digital neuromorphic recurrent neural network accelerators that exploit activation and weight sparsity to dramatically reduce memory access and speed up the inference. His accelerators achieved state of the art performance and efficiently scale to use economical DRAM memory. He developed variants that run on the cheapest FPGA development boards. He showed how they were capable of continuous automatic speech recognition and robotic control.

Gao developed this work during his PhD with the Sensors Group at the Institute of Neuroinformatics, University of Zurich and ETH Zurich. He is now a tenure-track assistant professor at TU Delft where he has started a lab entitled "Efficient circuits & systems for Machine Intelligence" (EMI). Gao's Google scholar profile provides links to his work.

Gao's PhD project contributions and publications
Gao demonstated how his first RNN accelerator "DeltaRNN" could accurately recognize speaker-independent, multi-accent digits during a live project meeting.
Gao's EdgeDRNN accelerator controlled Caltech's AMBER lab Rachel Gehlhar's AMPRO e-leg with performance indistinguishable to the PID trajectory controller, opening the possibility of future more flexible and adaptive neural control.

Gao's PhD work is described in a series of publications.

  1. D. Neil et al., "Delta Networks for Optimized Recurrent Network Computation," ICML, 2017
  2. C. Gao et al., "DeltaRNN: A Power-efficient Recurrent Neural Network Accelerator," FPGA, 2018
  3. C. Gao et al., "Real-time Speech Recognition for IoT Purpose using a Delta Recurrent Neural Network Accelerator," ISCAS 2019
  4. C. Gao et al., "Live Demonstration: Real-Time Spoken Digit Recognition Using The Deltarnn Accelerator," ISCAS, 2019
  5. C. Gao et al., "EdgeDRNN: Enabling Low-latency Recurrent Neural Network Edge Inference," AICAS, 2020 (Best Paper Award)
  6. C. Gao et al., "EdgeDRNN: Recurrent Neural Network Accelerator for Edge Inference," IEEE Journal on Emerging and Selected Topics in Circuits and Systems (JETCAS), 2020
  7. C. Gao*, R. Gehlhar*, et al., "Recurrent Neural Network Control of a Hybrid Dynamical Transfemoral Prosthesis with EdgeDRNN Accelerator," IEEE International Conference on Robotics and Automation (ICRA), 2020
  8. K. Kim*, C.Gao* et al., "A 23μW Solar-Powered Keyword-Spotting ASIC with Ring-Oscillator-Based Time-Domain Feature extraction," IEEE International Solid-State Circuits Conference (ISSCC), 2022
  9. K. Kim*, C.Gao et al., "A 23-μW Keyword Spotting IC With Ring-Oscillator-Based Time-Domain Feature Extraction," IEEE Journal of Solid-State Circuits (JSSC), 2022
  10. C. Gao et al., "Spartus: A 9.4 TOp/s FPGA-based LSTM Accelerator Exploiting Spatio-temporal Sparsity," IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2022