An integrated benchtop hyperspectral sensor and analysis system was developed for use in laboratory and field applications. This demonstrated the application of compact hyper-spectral camera system utilizing a liquid crystal tunable filter for ground truth collection and analysis in support for remote sensing applications. It also demonstrated a novel algorithmic paradigm for neural network training specifically optimized for computationally intensive applications such as hyperspectral analysis. During the project, a toolbox of additional algorithms for hyperspectral analysis was developed.

Cybernet performed research and developed a robotic testbed to demonstrate a system for using computer vision-based methods to solve docking problems during space missions. The vision system matched models from CAD sources to one or more images from calibrated CCD cameras.

Cybernet developed a detailed study showing the usefulness of adaptive network decision-making in real-world applications. The review included a consideration of advantages/disadvantages over traditional decision-making methods; techniques for training, testing and evaluating systems; and network functions. Adaptive networks which are covered include back propagation, counter propagation, bi-directional associative memory, Hopfield, brain-state-in-a-box, self-organizing map, perceptron, functional link, adaline, madaline, adaptive resonance theory 1, and genetic algorithm.

Cybernet developed an intelligent satellite telemetry diagnostic assistant for incorporation into future satellite ground stations. Based on neural net and adaptive artificial intelligence technology, the system provides anomaly/fault detection, isolation, and resolution in satellite telemetry streams. The system takes advantage of the ability of this advanced technology to continuously monitor complex data streams, track emerging trends, and maintain constant readiness for the onset of failure.

The use of robotically controlled excavation vehicles for defusing unexploded ordnance (UXO) was facilitated by Cybernet contributed machine recognition system that was integrated with an Air Force UXO excavator.. In this project, Cybernet performed a comprehensive study of existing defusing technology and UXO conditions, developed a database of ordnance data in modeling format, and developed proof-of-concept systems for a machine vision recognition system and an image/model-based pose determination system.

Automated digital video analysis of lab animal behaviors, especially rodents' behaviors. These behaviors include home cage activity, open field locomotion, object recognition, various mazes, fear conditioning, depression tests, etc.

INSL develops radio frequency coils integrated into animal holders for high field magnetic resonance imaging in conscious mice, rats, and monkeys.

The NIH SBIR/STTR program has significantly contributed to the overall technology development of IOSs portfolio of sensing, imaging, and toxin detection products. This in turn has helped the Company maintain a strong image, and has established IOSs reputation for leadership in advanced optical sensing and analytical technology development that continues to grow worldwide. The fully distributed DICAST sensing methodology bridges the gap between point detection and standoff detection, and is based on fiber optic cables in which the sensor material is the cladding component of the fiber optic itself, hence the entire length of the cable is agent sensitive. making it ideal for the protection of fixed assets.

The paper presents a formulation and an implementation of a system for voice output extraction (VOX) in real-time and near-real-time realistic real-world applications. A key component includes voice-signal separation and recovery from a mixture in practical environments. The signal separation and extraction component includes several algorithmic modules with a variety of sophistication levels, which include dynamic processing neural networks in tandem with (dynamic) adaptive methods.

Develop a visual object recognition architecture using Hidden Markov Models (HMM). Specifically, develop a lip reading prototype program that recognizes spoken digits (0-9) using only the video output (i.e., no audio). Combine the visual speech recognition technology with a commercially available audio based speech recognition engine to make a robust audio visual speech recognition (AVSR) system. The resulting AVSR system will have a much better accuracy than a conventional audio only speech recognition system.

IMAGE VUE is a wizard-driven software system, running under Windows XP, for co-registering EEGs with MRIs, performing patented DEBLURRING spatial enhancement and several types of source localization analysis, and interactive 3-D graphics visualization. The patented XCALIPER hardware and associated software facilitates rapid measurement of EEG electrode positions needed for co-registration with MRIs.

The SmartWheel examines the biomechanics of wheelchair propulsion by precisely measuring parameters including stroke frequency, propulsion angle, acceleration, forces applied to the handrim, velocity, and distance traveled. The SmartWheel is the only commercial product in the world that measures propulsion biomechanics in the natural environment of the wheelchair user. The SmartWheel's high-speed wireless technology, sophisticated on-board computer, and user-friendly interface allow for the collection of wheelchair propulsion data in nearly any lab, clinical or field setting. The unique capabilities that the SmartWheels offers have the potential to revolutionize wheelchair-related rehabilitation research and clinical practice.