Output details

Synthesis of nonseparable 3-D spatiotemporal bandpass filters on analog networks

Filter-based models of visual neurons are often constrained to employ separable functions of space and time, making synthesis easier. This paper was first to propose more general, non-separable functions, suggesting how they might be implemented in circuitry. It extends the available tools for designing spatio-temporal filters to cover non-separable cases. Early digital prototypes, and a methodology for achieving low-power implementations, were instrumental in winning seed investment for Cortexica, an Imperial spin-out (www.imperialinnovations.co.uk; 2009). Cortexica, in turn, supported and participates in a €5.3M, 4-year, FP7 Initial Training Network (#289146 “Neural Engineering Transformative Technologies (NETT)”) that funds 20 researchers in 5 countries.