Skip to Content

Mon, Oct 3

Speaker: Dr. Jana Andrejkova and Dr. Zlatica Dolna

Title: An Overview of Research in Biomedical and Rehabilitation Engineering at the Technical University of Kosice (TUKE), Slovakia

Abstract: This seminar will be presented in three parts. In the first part we will present a short introduction to the research programs, the faculty and the departments at the Technical University of Kosice (TUKE) in Kosice, Slovakia.  In the second part, Dr. Jana Andrejkova will present the results of her PhD work on Rehabilitation Engineering from the Assistive Technologies Point of View: Experimental Evaluation of e-services for Elderly and People with Impairment. This project is devoted to an experimental verification of e-services' applications for a target group that includes seniors and people with sensory impairments. The work provides a state-of-the-art review of selected e-services, information and communication technologies (ICT) and the evaluation of the accessibility for the target group. The theoretical background builds on knowledge of design, on the accessibility rules and standards, and on the application of statistical methods for data processing from end-users, from testing e-services and from ICTs. The practical part of the research is focused on testing the SSIS (Smart Speech Interactive System) by groups of people with and without some level of visual impairment;. This includes the testing of intelligent house appliances (e.g.,  refrigerator, dish washer, washing machine, and an air-conditioning unit linked to the internet); a blood pressure monitor connected to a PC, and the evaluation of the electronic accessibility of a website to the international project MonAMI, as well as testing of the technology NEMO, which enables the remote control of intelligent devices in the household environment using voice commands. All the testing was conducted by individuals with visual impairment. However, the accessibility evaluation is considered from the point of view of a user with other impairments and/or the elderly.  In the third part, Dr. Zlatica Dolna will present the results of her PhD work on Application of Human Motion Analysis in Biometrics: Measurement of Gait Parameters and Their Variability. This project is derived from research on human motion analysis and gait variability during the gait cycle. We used the human motion analysis system SMART from BTS, which is based on optical motion capture from passive markers attached to the body. A set of parameters that gives the most significant differences between people was derived for the motion pattern that represents individual features, including information about specific movement patterns or stereotypes.  Such information is characteristic for individuals and could be used to characterize disease, motor impairment or even human identification law enforcement purposes. The basic assumption for the analysis was that human gait parameters are relatively steady and that an invariant characteristic is unique for each person - as a gait finger print. In addition, it was assumed that there is a statistically significant difference between individuals when a comparison based on the motion features is conducted. Therefore, it is possible to find or recognize the identity of individuals using gait parameters that meet both conditions. This work introduces new measurement methodology and methods for motion data assessment as well as for their evaluation. The main goals were to assess gait parameters, to find a prototype gait pattern defined using experimental methods and to develop new mathematical-statistical techniques. The results, obtained under laboratory conditions, demonstrated the ability to record and monitor gait data, to evaluate individual gait patterns, and to define the motion stereotype. The results of this study provide a new biometric system that may be used outside the laboratory for human mobility identification and potentially in criminology for monitoring criminal activity.
Host: Dr. Patton
************************

Speaker: James Hedges, Ph.D.

Title: Integration in MT neurons and in the perception of visual motion

Abstract:  I will describe two projects on the perception and integration of visual motion. In the first project, we defined the spatial scale at which area MT neurons represent motion with anesthetized and awake monkey neurophysiological recordings, and we determined whether human perception is consistent with those representations with separate psychophysical experiments. Most MT neurons are strongly direction selective and have large receptive fields, built by combining inputs with smaller receptive fields, each of which responds to local motion. To assess whether MT neurons integrate these inputs in a way that allows them to represent global motion, we used a stimulus in which the directions of local and global motion are independently controlled. We found that MT responses depended only on local motion, and yet under similar conditions, human observers perceive global motion and are impaired in discriminating local motion. These results suggest that although the perception of some forms of local motion could plausibly depend on MT signals, the perception of global motion must involve mechanisms qualitatively different from those in MT. In the second project, we developed a model for how and why perceptual interpretations of visual motion are selected. Foundational to the second project was the observation that the local spatiotemporal pattern of light on the retina is often consistent with a single translational velocity, but may also be interpreted as a superposition of multiple spatial patterns drifting with different velocities. Human perception reflects both interpretations. In our model, an observer's percept corresponds to the most probable interpretation of noisy measurements of local image motion, based on separate prior beliefs about the speed and singularity of visual motion. Predictions from our model are well-matched to human perceptual interpretations across a broad range of stimulus conditions. Our model is generalizable and could be augmented to predict the perception of arbitrary spatiotemporal visual inputs.
Host: Dr. Kording