Shih-Chii Liu Shih-Chii Liu
Book: Event-Based Neuromorphic Systems (JUST appeared 2015)
by Liu, Delbrück, Indiveri, Whatley and Douglas. Publisher: Wiley UK.
Book: Analog VLSI: Circuits and Principles
by Liu, Kramer, Indiveri, Delbrück, and Douglas. Publisher: MIT Press. ISBN: 0-262-12255-3.
Open Projects (Semester Project or Master Project)
Look on the Sensors
group webpage for additional projects.
My research interests are in developing computational models of sensory and cortical processing; including spike-based models and algorithms suitable for implementation in hybrid silicon analog-digital systems. These computational principles can then be embedded in proto-products. Other groups within the Institute who work on neuromorphic circuits include those of Tobi Delbrück and Giacomo Indiveri .
The AEREAR2 cochlea is a binaural event-driven cochlea with 64 channels. It models the basilar membrane, inner hair cells, and spiral ganglion cells of the cochlea. Its use is being investigated in a number of applications including localization and sound recognition.
Spike timing information from the AEREAR2 can carry useful information for various auditory tasks. Applications of the silicon cochlea include auditory source localization harmonic sound detection, harmonic vs disharmonic sound recognition, and speaker verification.
How can information from different sensors be combined to provide coherence in our internal object representation? In this project, we use specifically the timing information carried in the outputs of both the AEREAR2 cochlea spikes and the visual spikes of the Dynamic Vision Sensor constructed by Lichtsteiner and Delbrück . We also investigate a framework for combining this information in a probabilistic network.
We have developed on a computational model of visual and auditory spatial map alignment in the inferior colliculus of the barnowl using STDP learning. This work is done together with Prashanth D'Souza and Richard Hahnloser. The results of this work will be applied towards the hardware sensors. Partially funded by the Zurich Neuroscience Center.
Neuromorphic motion chips can be useful for robotic platforms and microflyers because of their low-power and low-latency properties. The bio-inspired vision-based microflyer project investigates the use of such sensors on very light-weight flying platforms. Funded by the Swiss National Science Foundation.
Together with YingXue Wang, we are looking at active dendritic circuits and the computational role of nonlinearities in a neuronal model. These nonlinearities can be integrated together with short-term plastic synapses as part of a integrate-and-fire silicon neuron model so as to implement the putative processing of cortical cells. This research was reported in the 29th March 2001 issue of Nature. This report shows results from this cortical-like chip.
Together with Bradley A. Minch at Olin University, we are looking at ways of incorporating homeostatic, adaptation, and learning mechanisms in integrate-and-fire neurons using floating-gate technology.
We are developing the infrastructure for assembling reconfigurable asynchronous event-based multi-chip systems. Systems typically use spike-based front-end vision and/or audition sensors and neuronal networks for the subsequent computation. The chips in the system communicate using the address-event representation (AER) protocol.
An example system is CAVIAR which is the product of an EU-funded 5th FET framework project centered around the assembly of a spike-based asynchronous vision system. The partners in the projects are IMSE in Sevilla, Spain, UIO, and the University of Sevilla.
Email: shihatini.phys.ethz.ch