Research
Thrusts
.....Nanostructures
.....Biomimetics
.....Bio-Nano
Jose Principe - Research on time-carrying signal processing with neural networks focuses on the development of self-organizing and locally recurrent neural network models to process signals that vary over time (as the gamma model, which was invented at UF). The Computational NeuroEngineering Laboratory is also studying the performance of the gamma model for word spotting, which is the recognition of a key word in a stream of continuous speech.The CNEL has developed a new algorithm based on two dimensional gamma kernels (quadratic gamma detector) to discriminate targets from the conventional technique and it is presently included in the MIT/Lincoln Lab test facility. Extensions to UWB are currently being developed.The CNEL has started a major effort in auditory signal processing. Biologically-plausible models are used to understand how humans localize, process, and understand sounds. New neural network algorithms are used to cope with the cocktail party effect (the unscrambling of several speakers voices) and the automatic segmentation of sounds. The CNEL is studying the biological basis for the precedence effect. Ultimately these lines of research will result in algorithms that will be imbedded in special-purpose real-time sound/speech processing hardware.The CNEL is also studying biologically-inspired analog signal processing techniques in several application areas such as: 3-D sound localization and processing, analog speech synthesis, visual motion detection, wavelet image enhancement, adaptive filters, and chaotic oscillators. Ultimately this research will result in products for speech recognition preprocessors, hearing aids, teleconferencing, real-time image processing, and intelligent automobiles. This lab is also using analog VLSI to gain a better understanding of biological computation. Models of the cochlear and retinal processing have been successfully fabricated.