AFIT neural network research

A brief summary of research done at the Air Force Institute of Technology (AFIT) in the area of neural networks is provided. It has been shown that backpropagation, used for feedforward artificial neural networks, is just a degenerate version of an extended Kalman filter, and that networks can do about as well as the optimum statistical classification technique. A method of finding the importance of features for use by a neural network classifier has been determined. Techniques for using neural networks for image segmentation have been developed. In optical pattern recognition, techniques that allow the processing of real FLIR (forward-looking infrared) images with existing binary spatial light modulators have been devised. An optical direction of arrival detector applicable to laser illumination direction determination has been designed and tested; the design is similar to a fly's eye. Coated mirrors for the optical confocal Fabry-Perot interferometer have been designed, specified, fabricated, and installed. Significant progress has been made in the use of neural networks for processing multiple-feature sets for speech recognition     

A wavelet multiresolution analysis for spatio-temporal signals

The wavelet filters of the conventional 3D multiresolution analysis possess homogeneous spatial and temporal frequency characteristics which limits one's ability to match filter frequency characteristics to signal frequency behavior. Also, the conventional 3D multiresolution analysis employs an oct-tree decomposition structure which restricts the analysis of signal details to identical resolutions in space and time. This paper presents a 3D wavelet multiresolution analysis constructed from nonhomogeneous spatial and temporal filters, and an orthogonal sub-band coding scheme that decouples the spatial and temporal decomposition processes     

Temporal changes in fluid chemistry and energy profiles in the vulcano island hydrothermal system

In June 2003, the geochemical composition of geothermal fluids was determined at 9 sites in the Vulcano hydrothermal system, including sediment seeps, geothermal wells, and submarine vents. Compositional data were combined with standard state reaction properties to determine the overall Gibbs free energy (DeltaG(r) ) for 120 potential lithotrophic and heterotrophic reactions. Lithotrophic reactions in the H-O-N-S-C-Fe system were considered, and exergonic reactions yielded up to 120 kJ per mole of electrons transferred. The potential for heterotrophy was characterized by energy yields from the complete oxidation of 6 carboxylic acids- formic, acetic, propanoic, lactic, pyruvic, and succinic-with the following redox pairs: O(2)/H(2)O, SO(4) (2)/H(2)S, NO(3) ()/NH(4) (+), S(0)/H(2)S, and Fe(3)O(4)/Fe(2+). Heterotrophic reactions yielded 6-111 kJ/mol e(). Energy yields from both lithotrophic and heterotrophic reactions were highly dependent on the terminal electron acceptor (TEA); reactions with O(2) yielded the most energy, followed by those with NO(3) (), Fe(III), SO(4) (2), and S(0). When only reactions with complete TEA reduction were included, the exergonic lithotrophic reactions followed a similar electron tower. Spatial variability in DeltaG(r) was significant for iron redox reactions, owing largely to the wide range in Fe(2+) and H(+) concentrations. Energy yields were compared to those obtained for samples collected in June 2001. The temporal variations in geochemical composition and energy yields observed in the Vulcano hydrothermal system between 2001 and 2003 were moderate. The largest differences in DeltaG(r) over the 2 years were from iron redox reactions, due to temporal changes in the Fe(2+) and H(+) concentrations. The observed variations in fluid composition across the Vulcano hydrothermal system have the potential to influence not only microbial diversity but also the metabolic strategies of the resident microbial communities.     

An application of embedology to spatio-temporal pattern recognition

The theory of embedded time series is shown applicable for determining a reasonable lower bound on the length of test sequence required for accurate classification of moving objects. Sequentially recorded feature vectors of a moving object form a training trajectory in feature space. Each of the sequences of feature vector components is a time series, and under certain conditions, each of these time series has approximately the same fractal dimension. The embedding theorem may be applied to this fractal dimension to establish a sufficient number of observations to determine the feature space trajectory of the object. It is argued that this number is a reasonable lower bound on test sequence length for use in object classification. Experiments with data corresponding to five military vehicles (observed following a projected Lorenz trajectory on a viewing sphere) show that this bound is indeed adequate     

Efficient numerical algorithm for steady-state Kalman covariance

A stable, quadratically convergent numerical algorithm is presented for computing the steady-state covariance and gain matrices of the Kalman filter. The method is more rapidly convergent than standard Riccati integration techniques and is easier to implement than existing eigenvalue-eigenvector algorithms. The quadratic convergence is proved analytically and illustrated by a numerical example     

Magnifying our world: Why we must extend civilization to the Moon

Since the completion of its original prehistoric migration some 10 000 years ago, humankind has had nowhere else on Earth to go. The Space Age has produced the wherewithal for a new off-Earth human migration to restart, with the Moon as the clear initial destination. As insurance against cataclysmic and human-induced natural disasters, as a means of creating new commercial industrial–financial opportunities at a time of fundamental change in the global economy, and as a way for the USA to regain some international political influence, it should pursue Moon settlement vigorously and soon. However, the developing US space exploration program being designed by NASA for such an undertaking is too narrowly confined to the pursuit of natural-science objectives. There must be fundamental involvement of the private sector and of a much wider range of professionals and federal departments therein.     

Reconstruction of mutilated speech

A system has been developed to enhance the quality of mutilated speech. A standard spectrogram analysis of the damaged speech is performed. The speech is then resynthesized as a sum of computer-generated sinusoids whose amplitudes and phases are derived partly from the given mutilated speech signal and partly from rules based on known properties of normal speech. The sinusoids selected are only approximate harmonics of the glottal pitch and are selected by a nonlinear, noncausal set of rules to reduce the nonspeech components in the synthesized speech output. The system has been shown to increase the quality of the mutilated speech appreciably     

Alpha-Beta Filter with Correlated Measurement Noise

For many tracking applications, the measurement errors onsuccessive observations are correlated. Using a first-order Markov model for the correlation, we present analytical expressions for the time-varying covariance and gains of an alpha-beta tracking filter.To a good approximation, the effect of correlation is to increase the time interval between measurements by a factor (1+a)/(1-a),where a is the coefficient of correlation between successive measurements.