Face recognition: the problem of compensating for changes inillumination direction

A face recognition system must recognize a face from a novel image despite the variations between images of the same face. A common approach to overcoming image variations because of changes in the illumination conditions is to use image representations that are relatively insensitive to these variations. Examples of such representations are edge maps, image intensity derivatives, and images convolved with 2D Gabor-like filters. Here we present an empirical study that evaluates the sensitivity of these representations to changes in illumination, as well as viewpoint and facial expression. Our findings indicated that none of the representations considered is sufficient by itself to overcome image variations because of a change in the direction of illumination. Similar results were obtained for changes due to viewpoint and expression. Image representations that emphasized the horizontal features were found to be less sensitive to changes in the direction of illumination. However, systems based only on such representations failed to recognize up to 20 percent of the faces in our database. Humans performed considerably better under the same conditions. We discuss possible reasons for this superiority and alternative methods for overcoming illumination effects in recognition     

Monitoring of antibiotic use in a primary and tertiary care hospital

Prophylactic and curative use of antibiotics was studied prospectively in 87 consecutive medical and surgical cases of a tertiary care hospital and in 98 cases of a primary care hospital. Based on Kunins' criteria, antibiotic prophylaxis was found to be more inappropriate in the primary care hospital (49%) than in the tertiary care hospital (34%). Antibiotic therapy, however, was more appropriate at the primary level; 67% as opposed to 60% at the tertiary level. This resulted in a similar overall level of inappropriate antibiotic use in the two hospitals. Surgical prophylaxis was started postoperatively in 68% of the primary care hospital cases. Though prophylaxis was always perioperative in the tertiary care hospital, the postoperative duration was more than 7 days in one third of the cases. The nosocomial infection rate in those given prolonged prophylaxis was higher than those who received antibiotics for less than 72 hours. Antibiotics were started empirically in 78% of tertiary hospital care cases and 100% of cases in the primary hospital. Though culture sensitivity was done in 80% of the tertiary care cases, more than half the specimens were sent after multiple doses of antibiotics were started. The choice of antibiotic did not always correlate with the sensitivity report. Though cost-effective drugs were chosen in 50% of cases, in more than 20% of cases expensive drugs were started. The study highlights the need for an antibiotic audit and suggests the necessity of having an ongoing peer audit.     

Importance of rotational losses in rotating machines and transformers

There are increasing calls for a standard specification for the measurement of rotational losses in electrical steels and amorphous magnetic materials. However, the question still remains as to whether rotational losses are sufficiently important in practice to merit the implementation of a special standard. This paper explains what rotational magnetization is and how the losses occur. It then gives a brief explanation of the measurement difficulties encountered. Regions of transformers and rotating machine cores where rotational flux occur are illustrated and an estimate of the resulting rotational losses is made. It is shown that the proportion of rotational loss in transformers is not generally high, whereas in motors it can amount to more than 50% of the total core loss. The results indicate that rotational loss is of sufficient importance in motors to merit the establishment of standardized methods of measuring it in magnetic materials.     

Electroabsorption studies of undoped C60 thin films

Electroabsorption spectra have been measured for undoped C₆₀ thin films at 80 and 300 K in the spectral range 1.76 to 5 eV. Absorption and thermal modulation spectra have also been obtained. The electronic energy levels of C₆₀ are reviewed, and the electroabsorption spectra are discussed in terms of the Stark effect on energy levels of A_g, H_g and T_u symmetries.     

The regulation of neovascularization of matrix metalloproteinases and their inhibitors

This study sought to produce a cognitive profile of herpes simplex encephalitis (HSE) survivors from a large group of definitively diagnosed, acyclovir-treated participants. Results from 22 adults who underwent a battery of neuropsychological tests indicated anterograde memory dysfunction to be the most severe and common deficit (although the variation was great), with less severe and less frequent impairments in the areas of retrograde memory, executive functions, and language functioning. Overall, neuropsychological outcome was unimpaired in six participants, mildly impaired in thirteen, moderately impaired in one, severely impaired in two. Older participants and those with a lower level of consciousness before the start of treatment produced poorer scores on certain aspects of cognitive outcome (p < 0.05). A significantly better cognitive outcome was found in participants for whom there was a short delay (fewer than 5 days) between symptom onset and acyclovir treatment compared with those participants for whom there was a longer delay. The two children in the study had disparate results on most tests, the exception being those assessing memory functioning on which both children had scores at population norms. On a naming task designed to explore category-specific knowledge deficits, the adults as a group made more errors on pictures of living things than nonliving things (matched pair-wise for word frequency and visual familiarity), although this difference disappeared on a smaller subset of pictures also matched for visual complexity.     

Resonant Raman scattering study of undoped fullerene thin films

Raman spectra of the fullerenes C₆₀ and C₇₀ have been measured for photon energies from 2.1 to 3.6 eV. Resonant enhancement of the Raman cross section versus wavelength has been measured for the 270, 497, 1424, 1469, and 1568 cm⁻¹ lines, with the diamond 1332 cm⁻¹ line as a reference. A prominent enhancement occurs near 2.4 and 2.5 eV for the C₆₀ 497 cm⁻¹ and 1469 cm⁻¹ lines, respectively. The C₇₀ 1424 cm⁻¹ and 1568 cm⁻¹ lines increase monotonically in intensity between 2.3 and 2.7 eV. The 270 cm⁻¹ line, which has contributions from both C₆₀ and C₇₀, exhibits resonance enhancement which is qualitatively similar to that of the C₇₀ lines. All of the lines show resonance enhancement at 3.4 and 3.5 eV.     

Programming simultaneous actions using common knowledge

This work applies the theory of knowledge in distributed systems to the design of efficient fault-tolerant protocols. We define a large class of problems requiring coordinated, simultaneous action in synchronous systems, and give a method of transforming specifications of such problems into protocols that areoptimal in all runs: these protocols are guaranteed to perform the simultaneous actions as soon as any other protocol could possibly perform them, given the input to the system and faulty processor behavior. This transformation is performed in two steps. In the first step we extract, directly from the problem specification, a high-level protocol programmed using explicit tests for common knowledge. In the second step we carefully analyze when facts become common knowledge, thereby providing a method of efficiently implementing these protocols in many variants of the omissions failure model. In the generalized omissions model, however, our analysis shows that testing for common knowledge is NP-hard. Given the close correspondence between common knowledge and simultaneous actions, we are able to show that no optimal protocol for any such problem can be computationally efficient in this model. The analysis in this paper exposes many subtle differences between the failure models, including the precise point at which this gap in complexity occurs.This research was supported by the Office of Naval Research under contract N00014-85-K-0168, by the Office of Army Research under contract DAAG29-84-K-0058, by the National Science Foundation under Grant DCR-8302391, and by the Defense Advanced Research Projects agency (DARPA) under contract N00014-83-K-0125, and was performed while both authors were at MIT. A preliminary version of this work appeared in theProceedings of the 27th Annual IEEE Symposium on Foundations of Computer Science, Toronto, 1986.This author was primarily supported by an IBM postdoctoral fellowship.     

Lake bathymetry from Indian Remote Sensing (P6‐LISS III) satellite imagery using artificial neural network model

The remote sensing technique provides a rapid and relatively inexpensive means of identifying silted areas in large water bodies, in order that desilting activities can be effectively conducted. This study developed lake bathymetry for a selected lake system (Akkulam–Veli Lake, Kerala, India) from the Indian Remote Sensing (IRS P6‐LISS III) satellite imagery, using an artificial neural network (ANN) model. The water depth was measured for 17 months at different points in the lake on the same date of overpass of the IRS satellite. The satellite imageries obtained for 12 December 2007 and 16 February 2009 were identified as cloud‐free images. ANN models were developed with the four input series of radiance values from green, red, NIR and MIR bands observed for the satellite imagery obtained on 12 December 2007 at the sampling sites, with actual water depth measurements also being taken on the same date. A three‐layered feed forward neural network with back propagation training algorithm was developed for this study. To train the model, it was run several times by changing the number of neurons, learning rate and the momentum constants until the mean square error was minimum. When the number of neurons is increased to 35, and the logsig function is used as ANN transfer function, the error becomes minimum. To test the model, the developed ANN was run for a new set of input from the satellite imagery taken on 16 February 2009. Comparing the predicted and measured values for the same sites for the same day, it was found that the model is best suited for predicting water depth using ANN and the radiance values for four bands of IRS satellite imagery. The results of this study indicated that, for the shallow lake with lower depth, the difference between the actual and predicted value was considerable. In contrast, this was not the case where the lake water depth was greater, indicating an increased prediction accuracy with ANN with increasing depths for shallow lakes. A bathymetry map prepared with ANN indicated only the lake shoreline, as well as the shallow littoral zones. The approach used in this study requires further refinement, including further of the model based on using more field measurements to obtain a better bathymetry map.     

Programming simultaneous actions using common knowledge

This work applies the theory of knowledge in distributed systems to the design of efficient fault-tolerant protocols. We define a large class of problems requiring coordinated, simultaneous action in synchronous systems, and give a method of transforming specifications of such problems into protocols that areoptimal in all runs: these protocols are guaranteed to perform the simultaneous actions as soon as any other protocol could possibly perform them, given the input to the system and faulty processor behavior. This transformation is performed in two steps. In the first step we extract, directly from the problem specification, a high-level protocol programmed using explicit tests for common knowledge. In the second step we carefully analyze when facts become common knowledge, thereby providing a method of efficiently implementing these protocols in many variants of the omissions failure model. In the generalized omissions model, however, our analysis shows that testing for common knowledge is NP-hard. Given the close correspondence between common knowledge and simultaneous actions, we are able to show that no optimal protocol for any such problem can be computationally efficient in this model. The analysis in this paper exposes many subtle differences between the failure models, including the precise point at which this gap in complexity occurs.