Enabled by the availability of both excitatory and inhibitory post-synaptic currents, a biological neural network is inherently capable of implementing more sophisticated non-monotonic classification schemes. While such currents are readily emulated in a software-based artificial neural network using both positive and negative synaptic weighting factors, the same is not as straight forward in a hardware implementation. In this work, two dual-gate thin-film transistors with effectively infinite direct-current input impedance are deployed to construct the excitatory and inhibitory conduction channels of an artificial synapse. The utility of a hardware-based artificial neural network constructed using such synapses is demonstrated by its deployment in the implementation of the complete set of sixteen 2-input binary logic functions exhibiting both monotonic and non-monotonic behavior. The set of weighting factors needed for the implementation of each function are determined using a neuromorphic feed-forward training algorithm based on gradient-descent. While some of the functions can be implemented using the simplest reconfigurable$\mathbf {2\times 1}$network, all 16 of the functions can be implemented using a deeper, reconfigurable$\mathbf {3\times 2\times 1}$network.

Reported presently are two designs to improve the performance of a “chevron” electro-thermal in-plane actuator. One incorporates beams with uniform cross-sections but nonuniform lengths and tilt angles to accommodate the thermally induced expansion of the “shuttle”; the other incorporates beams with non-uniform cross-sections to achieve a wider spread of the high temperature “expansion” regions of the beams. With the product of the actuation force and displacement defined as a figure-of-merit, it is verified using finite-element simulation that the incorporation of non-uniform lengths and tilt angles, and non-uniform beam cross-sections leads to respective improvement of 10 and 65% in the figure-of-merit. The effectiveness of these designs was also tested by micro-fabricating actuators occupying fixed device areas.

Conventional acoustic microphones have limited bandwidth, hence inadequate for measuring aero-acoustic waves that require a wide bandwidth from hundreds of kHz to a few MHz and a large pressure dynamic range from ~10Pa to ~1kPa. A lumped-element model is developed and used to design and estimate the static and dynamic mechanical response of an aeroacoustic microphone. Also reported is the implementation and preliminary characterization of such a microphone with a diaphragm supported at the corners by cantilever springs and integrated with piezoresistive transducers made of metal-induced laterally crystallized polycrystalline silicon. A static sensitivity of ~0.1μV/V/Pa and a resonant frequency of 560kHz have been measured.

Fractal image encoding is a computationally intensive method of compression due to its need to find the best match between image subblocks by repeatedly searching a large virtual codebook constructed from the image under compression. One of the most innovative and promising approaches to speed up the encoding is to convert the range-domain block matching problem to a nearest neighbor search problem. This paper presents an improved formulation of approximate nearest neighbor search based on orthogonal projection and pre-quantization of the fractal transform parameters. Furthermore, an optimal adaptive scheme is derived for the approximate search parameter to further enhance the performance of the new algorithm. Experimental results showed that our new technique is able to improve both the fidelity and compression ratio, while significantly reduce memory requirement and encoding time.