We present the theory for heteroscedastic discriminant analysis (HDA), a model-based generalization of linear (LDA) derived in maximum-likelihood framework to handle heteroscedastic-unequal variance-classifier models. show how estimate Gaussian model parameters jointly with dimensionality reducing transform, using EM algorithm. In doing so, we alleviate need an priori ad hoc class assignment. apply theoretical results problem speech recognition and observe word-error reduction systems that...

An overview of the current-mode approach for designing analog VLSI neural systems in subthreshold CMOS technology is presented. Emphasis given to design techniques at device level using current-controlled current conveyor and translinear principle. Circuits associative memory silicon retina are used as examples. The methodology how it relates actual biological microcircuits discussed.

Polarization is a general descriptor of light and contains information about reflecting objects that traditional intensity-based sensors ignore. Difficult computer vision tasks such as image segmentation object orientation are made tractable with polarization techniques. Specularities, occluding contours, material properties can be readily extracted if the Stokes parameters available. Astrophysicists employ to measure spatial distribution magnetic fields on surface Sun. In medical field,...

The technical and economic benefits of artificial intelligence (AI) are counterbalanced by legal, social ethical issues. It is challenging to conceptually capture empirically measure both downsides. We therefore provide an account the findings implications a multi-dimensional study AI, comprising 10 case studies, five scenarios, impact analysis human rights AI known potential threats vulnerabilities. Based on our findings, we separate ethics discourse into three streams: (1) specific issues...

Accelerometers are common on many devices, including those required for text-entry.We investigate how to enter text with devices that solely enabled accelerometers.The challenge of text-entry such can be overcome by the careful investigation human limitations in gestural movements accelerometers.Preliminary studies provide insight into two potential designs purely use accelerometers gesture recognition.In experiments, we evaluate effectiveness each designs.The first experiment involves...

Recently, polarization vision has been shown to simplify some important image understanding tasks that can be more difficult perform with intensity alone. This, together the general capabilities of for understanding, motivates building camera sensors automatically sense and process information. Described in this paper are a variety designs have built partial linearly polarized light, computationally sensed information at pixel resolution produce visualization reflected from scene, and/or...

We report on inter-chip bidirectional communication and power transfer between two stacked chips. The experimental prototype system components were fabricated in a 0.5-mum silicon-on-sapphire CMOS technology. Bi-directional the chips is experimentally measured at 1Hz-15 MHz. circuits floating top chip are powered with capacitively coupled energy using charge pump. This first demonstration of simultaneous nongalvanic data stack. potential use 3-D VLSI aimed reducing costs complexity that...

Vertical p-n junction organic-inorganic hybrid diodes are fabricated and characterized. The current density rectification speed in the open atmosphere higher than previously reported organic vertical with relatively lower operating voltage, stability is encouraging.

We discuss the design, fabrication, and testing of a hybrid microsystem for stand-alone cell culture incubation. The micro-incubator is engineered through integration silicon CMOS die heater temperature sensor, with multilayer silicone (PDMS) structures namely, fluidic channels 1.5-mm diameter 12-muL well. A 90-mum-thick PDMS membrane covers top well, acting as barrier to contaminants while at same time allowing cells breath exchange gases ambient environment. packaging includes flexible...

In this paper, we present an architecture and corresponding analysis for large-scale neuromorphic systems using a digital approach where neurons are abstracted as arithmetic logic units communication processors. After presenting the architecture, establish few basic architectural principles, particularly scaling of system to large arrays. We demonstrate reality single chip million neuron built from these principles in commercial off shelf FPGA.

Abstract N , ′‐bis(3‐(perfluoroctyl)propyl)‐1,4,5,8‐naphthalenetetracarboxylic acid diimide (8–3‐NTCDI) was newly synthesized, as were related fluorooctylalkyl‐NTCDIs and alkyl‐NTCDIs. The 8–3‐NTCDI‐based organic thin‐film transistor (OTFT) on an octadecyltrimethoxysilane (OTS)‐treated Si/SiO 2 substrate shows apparent electron mobility approaching 0.7 cm V ‐1 s in air. fluorooctylethyl‐NTCDI (8–2‐NTCDI) fluorooctylbutyl‐NTCDI (8–4‐NTCDI) had significantly inferior properties even though...

Beginning with Amdahl's law, we derive a general objective function that links parallel processing performance gains at the system level, to energy and delay in subsystem microarchitecture structures. The employs parameterized models of computation communication represent characteristics processors, memories, communications networks. interaction latter microarchitectural elements defines global terms energy-delay cost. Following derivation, demonstrate its utility by applying it problem Chip...

Abstract A miniature atomic scalar magnetometer based on the rubidium isotope 87 Rb was developed for operation in space. The instrument design implements both M x and z mode leverages a novel microelectromechanical system (MEMS) fabricated vapor cell custom silicon‐on‐sapphire (SOS) complementary metal‐oxide‐semiconductor (CMOS) integrated circuit. has volume of only 1 mm 3 so that it can be efficiently heated to its operating temperature by specially designed, low‐magnetic‐field‐generating...

We present a fully automated system which unites CCD camera technology with liquid crystal to create polarization capable of sensing the partial linear reflected light from objects at pixel resolution. As not only measures intensity but also additional physical parameters light, it can therefore provide richer set descriptive constraints for understanding images. Previously has been shown that cues be used perform dielectric/metal material identification, specular and diffuse reflection...

Foreword. Preface. Acknowledgments. Contributors. Introduction (E. Sanchez--Sinencio). A Current--Based MOSFET Model for Integrated Circuit Design (C. Montoro, et al.). Review of the Performance Available Components Under Constraints Low--Power Operation (D. Bowers). Exploiting Device Physics in Efficient Computational Functions Analog VLSI (A. Andreou). Low--Voltage Techniques Using Floating--Gate Transistors (Chong--Gun Yu and Randall Geiger). CMOS Digital Circuits (S. Embabi). BiCMOS...

A VLSI architecture is proposed for the realization of real-time two-dimensional (2-D) image filtering in an address-event-representation (AER) vision system. The capable implementing any convolutional kernel F(x,y) as long it decomposable into x-axis and y-axis components, i.e., F(x,y)=H(x)V(y), some rotated coordinate system {x,y} if this product can be approximated safely by a signed minimum operation. intended to used complete system, known boundary contour feature (BCS-FCS) model,...

The performance of speech recognition systems is often improved by accounting explicitly for sources variability in the data. In SWITCHBOARD corpus, studied during 1994 CAIP workshop [Frontiers Speech Processing Workshop II, (August 1994)], an attempt was made to compensate systematic due different vocal tract lengths various speakers. method found a maximum probability parameter each speaker which mapped acoustic model mean models taken from homogeneous population. underlying that straight...

This paper revises the conditions under which translinear principle can be fully exploited for MOS transistors operating in subthreshold. Due to exponential nature of subthreshold transistors, applies immediately as long source-to-bulk voltages are made equal zero (or constant). addresses still satisfy a principle, but without imposing this constraint on all V/sub BS/ voltages. It is found that results more general formulation than originally BJT's since now multiple loops involved. The an...