Neuro Prosthetics and Brain Machine Interface
About
Our research focuses on the following areas: 1) Brain-machine interface: The grand goal is to use the neural interface to control and sense the interactions with dexterous upper limb prosthesis. The signals used are EEG, Electrocorticogram, and neural action potentials. 2) Motor: We are developing algorithms for control of the dexterous prosthetic hand using advanced machine learning (e.g. sparse coding) and deep learning (temporal neural networks) for robust, position/load independent dexterous control. 3) Sensory: We design sensors that mimic human skin to provide a sense of touch, pressure, temperature and pain perception and provide electrocutaneous and vibration feedback information to the amputee. The sensory percepts are analyzed through both behavioral studies as well as brain signal (EEG) mapping and localizing the sources.
Project Description
1. Brain-Machine Interface
The over-arching goal of upper limb prosthesis control is to provide intuitive, brain/mind controlled prosthesis. To reach this end, the central challenge is interpreting brain signals to derive the control signals from the brain. The key signals are Electroencephalogram (EEG), Electrocorticogram (ECoG), neural action potentials (spikes, local field potentials). We have applied each of these signal techniques, each with its unique advantage and limitation, for demonstrating dexterous control in monkeys as well as humans. The research pertained to neural signal analysis, network theory, decoding spikes/LFPs, and forming neural network and machine learning algorithms to interpret the patterns that decode for motor tasks.
2. Motor
a) We are developing algorithms for control of the dexterous prosthetic hand. After years of work on building brain-machine interface using EEG, ECoG, we are now back to the practical, affordable, non-surgical myoelectric EMG control. We use advanced machine learning (e.g. sparse coding) and deep learning (temporal neural networks) for robust, position/load independent dexterous control. B) To facilitate linking motor functions to underlying motor systems of the limb, we are building a biomechanical and motor/sensory embodiment model. C) To train an amputee, possibly at home, in the use of the dexterous hand, we employ augmented reality, HoloLens, and inertial unit technology – the ”Buzz-band”.
*replace with real image*
3. Sensory
An amputee also needs to feel and receive the sense of touch and corresponding embodiment of the sensory perception. Therefore, we design sensors that mimic human skin to provide a sense of touch, pressure, temperature and pain perception and provide electrocutaneous and vibration feedback information to the amputee. The se nsory percepts are analyzed through both behavioral studies as well as brain signal (EEG) mapping and localizing the sources. Our goal is to demonstrate the sensory perception not only enhances the quality of sensory perception, but practically interact with objects in daily lives, and also to show improvements in the control of the prosthesis.
