FAST FEATURE RANKING AND ITS APPLICATION TO FACE RECOGNITION

A fast feature ranking algorithm for classification in the presence of high dimensionahty and small sample size is proposed. The basic idea is that the important features force the data points of the same class to maintain their intrinsic neighbor relations, whereas neighboring points of different classes are no longer to stick to one an- other. Applying this assumption, an optimization problem weighting each feature is derived. The algorithm does not involve the dense matrix eigen-decomposition which can be computationally expensive in time. Extensive exper- iments are conducted to validate the significance of selected features using the Yale, Extended YaleB and PIE data- sets. The thorough evaluation shows that, using one-nearest neighbor classifier, the recognition rates using 100-- 500 leading features selected by the algorithm distinctively outperform those with features selected by the baseline feature selection algorithms, while using support vector machine features selected by the algorithm show less prominent improvement. Moreover, the experiments demonstrate that the proposed algorithm is particularly effi- cient for multi-class face recognition problem.     

The concept of selecting the main parameters of a lightweight airplane with regard to the requirements of the contemporary Russian market

The basic principles of selecting some design parameters of a lightweight airplane for the contemporary Russian market are presented. Recommendations for selecting these parameters with the service requirements specified are given based on the study performed.     

Spectres auroraux

Summary The paper presents a review of the main spectroscopic studies of polar aurora accomplished during the past eight years.In a first part, the few new results concerning knowledge of the spectrum itself are examined. They concern detailed spectrum analysis, a few spectral features (sodium doublet, helium lines) and the extension of spectral observations to very low wavelengths (3000 to 1000 Å). The main features discovered in this region are the Vegard-Kaplan and Lyman-Birge-Hopfield systems as well as the 1304 oxygen line. A comparison of observed and theoretical intensities shows serious discrepancies, not all of which may be attributed to difficulties of observation. Next is given classification, from a spectral point of view, of the various kinds of auroral phenomena, with respect to altitude and latitude. Particular attention is given to high altitude and sunlit aurorae, and to low intensity, non discrete aurorae observed around the 70th degree (geomagnetic) and around the polar cap.A second part underlines the influence of two important parameters, time and location, mainly as a possible means of detection of different mechanisms.As far as the time parameter is concerned, a few pages are devoted to statistical analysis of the great wealth of observational data accumulated over twenty or thirty years.It is shown that important conclusions emerge from such studies. Very small scale time variations are mainly concerned with the study of metastable states, their lifetimes and de-excitation due to shocks between particles. Differences of spectra with the location of the observed emitting volume is a much more recent subject, which has, however, produced important results. These concern mainly observations made from rockets, although a few results have been obtained from the ground. All these indicate significant differences that may shed some light on the difficult problem of mechanisms. The latter has not been developed, since a recent review has been written on the subject. The same is true for the very important problem of hydrogen emissions.The last pages are concerned with the latest developments of experimental methods, with some stress on image tubes, which will probably be one of the main detecting devices for observing auroral spectra in the near future.     

Accuracy of Star Occultation Measurements for Satellite Navigation

To determine the orbital parameters of a satellite using star occultations, it is necessary to measure the intensity of stars as they occult. The accuracy of the intensity measurements is reduced by interfering sources of light and internal noise, and is expressed in terms of the magnitude of various sources of interference. Subsequently, a specific system is analyzed in detail. Moonlit cloud tops introduce the largest systematic error. Photon noise from the star radiation, airglow, and moonlit cloud tops are the largest sources of random errors.     

Controlling spacecraft by means of solar radiation

The possibility of using solar sails in Earth orbit is investigated. The characteristic parameters of a solar sail consisting of six spheres attached by mutually orthogonal long rods are estimated.     

Toward Accurate On-Ground Attitude Determination for the Gaia Spacecraft

The work presented in this paper concerns the accurate On-Ground Attitude (OGA) reconstruction for the astrometry spacecraft Gaia in the presence of disturbance and of control torques acting on the spacecraft. The reconstruction of the expected environmental torques which influence the spacecraft dynamics will be also investigated. The telemetry data from the spacecraft will include the on-board real-time attitude, which is of order of several arcsec. This raw attitude is the starting point for the further attitude reconstruction. The OGA will use the inputs from the field coordinates of known stars (attitude stars) and also the field coordinate differences of objects on the Sky Mapper (SM) and Astrometric Field (AF) payload instruments to improve this raw attitude. The on-board attitude determination uses a Kalman Filter (KF) to minimize the attitude errors and produce a more accurate attitude estimation than the pure star tracker measurement. Therefore the first approach for the OGA will be an adapted version of KF. Furthermore, we will design a batch least squares algorithm to investigate how to obtain a more accurate OGA estimation. Finally, a comparison between these different attitude determination techniques in terms of accuracy, robustness, speed and memory required will be evaluated in order to choose the best attitude algorithm for the OGA. The expected resulting accuracy for the OGA determination will be on the order of milli-arcsec.     

Estimation of Correlation Functions by Stochastic Approximation

Two techniques for estimating the correlation function of a stationary random process using the method of stochastic approximation are developed. Both provide updated estimates as successive T-second observations are processed. Both assume a functional form for the correlation function that depends on a number of parameters that are estimated. One technique is based on the mean-square error; the other maximizes a likelihood function. Examples are included.     

Emesis and space motion sickness in amphibians.

Amphibians possess the ability to vomit in response to a variety of stimuli that provoke emesis in mammals. Pharmacological studies have establish that the ejection of gastric contents and the basic mechanism for vomiting have been phylogenetically conserved among these tetrapods. As part of on-going comparative studies on emesis in vertebrates, we previously documented that some postmetamorphic anurans and salamander larvae experience motion-induced emesis when exposed to the provocative stimulus of parabolic aircraft flight. However, more recent experiments suggest that there are strict conditions for inducing emesis in amphibians exposed to parabolic flight and that amphibians are not as sensitive to this stimulus as mammals. Further studies on emesis in lower vertebrates may help us understand the processes that cause emesis in abnormal gravitational regimes.     

Integration test project of CEEF--a test bed for Closed Ecological Life Support Systems.

CEEF (Closed Ecology Experiment Facilities) were installed at Rokkasho village in northern Japan, for the purpose of clarifying life-support mechanisms in a completely closed space, such as a Lunar or Mars base. An integration test using the Closed Plantation Experiment Facility and Closed Animal Breeding & Habitation Experiment Facility is needed before conducting an entire closed experiment including plants, animals and humans. These integration tests are planned to be conducted step by step from fiscal 2001 to 2008.