Are you interested in working on real-time novel spike-based algorithms based on event-driven sensors? We have a state-of-art silicon cochlea system AEREAR2.
The student will have an opportunity to improve and innovate on new audio algorithms and deep network solutions such as Deep Belief Networks based on this new output representation. We have example algorithms for source localization and source recognition. We also have a project to perform audio reconstruction from the spikes.
Real-time implementation of event-driven algorithms for sensory fusion
Interested in implementing real-time hardware systems for event-driven sensory fusion? We have a a state-of-art VLSI cochlea chip AEREAR2 and the Dynamic Vision Sensor (DVS) that produces event-driven outputs.
The student will have an opportunity to implement novel event-driven sensory-fusion algorithms both in software and on existing hardware platforms such as FPGAs.
Dendritic modeling and circuits
Active dendrites can have increased computational capabilities over passive dendrites. This project explores the properties of active dendrites with possible computational units to be implemented in VLSI circuits.
The student will develop simulations for studying the computational capabilities of dendrites with nonlinear properties.
Simulations will be done in Matlab/C++ but with a focus
on dendritic features for silicon circuits.
These features will be incorporated into dendritic circuits described in Wang and Liu, TCAS, 2011; Neural Computation 2010.
Learning of the synapses will also be considered (see Hussain et al, IJCNN 2013).
Active sensory perception and multi-modal integration
We have a robot head setup which holds the spike-based dynamic vision system (DVS)
and microphones connected to the AEREAR2 auditory sensor.
This project centers around multi-modal fusion models
that combine the spike outputs of these sensors for an active perceptual task.
He/she will develop the algorithms and motor control using jAER , a JAVA-based project for developing event-based algorithms.
On-board motion system for miniature flying platforms
The project looks the use of motion information computed from a neuromorphic retina sensor in a sensory-motor task. Low latency, low weight, and low power are key ingredients of the new aVLSI motion system. The student will learn how to build a new motion system on a small printed circuit board (PCB). The design will be based on a previous PCB design, and the student will have a chance to run experiments later when the printed circuit board is mounted on a flying platform.
Visit here for more information.
Building aVLSI circuits
Interested in building your own aVLSI circuits? We have on-going
design projects in VLSI design of biophysical
models of cochleas, spiking neuron circuits, and
motion circuits. You will have a chance
to design a circuit, and to test the circuit
once it is back from fab.
Or come by and talk to me or send email.
Look on the institute's project web site for