Ben H. Jansen, Ph.D., Professor, Electrical and Computer Engineering, UH
Understanding the physiological mechanisms that underlay neurosensory adaptation processes associated with spaceflight is essential to the development and evaluation of countermeasures to optimize crew health, safety, and performance. Following space flight, the ability to locomote is often severely impaired, which debility would jeopardize a crewmember's safety in an emergency. Understanding the underlying mechanisms associated with postflight locomotion decrements can directly lead to new or improved countermeasures. Therefore, we are developing software tools to quantify the changes observed in the gait patterns of astronauts returning from space flight.
Electromyographic (EMG) data on the electrical activity generated by muscles have been collected from four muscles in the left leg of the astronauts and the same muscles in the right leg, while walking on a treadmill, before and after spaceflight. Typically, data from about twenty contiguous steps are collected from an astronaut per session, with at least two sessions scheduled before spaceflight and two to three sessions afterwards. These noisylooking signals are digitized, rectified, and lowpass-filtered to obtain estimates of the amplitude envelope. A typical example is shown in Fig. 1.
Figure 1. Rectified and low-pass filtered EMG of the Rectus Femoris, Biceps Femoris, Tibialis Anterior, and Gastrocnemius recorded during one gait cycle from an astronaut.
The next stage of the processing is based on concepts derived from nonlinear dynamical system (chaos) theory and involves "trajectory-based" clustering. Trajectories are multidimensional displays of the evolution of the state variables of the system under study. In our case, the EMG collected from astronauts is used to construct a trajectory for each single stride. The trajectories provide a view of the phasing and activation levels of all the muscles simultaneously.
We are currently exploring a distance-based similarity measure to quantify the differences between trajectories obtained from an astronaut before and after spaceflight, and differences among astronauts. Specifically, we are using a hierarchical clustering procedure to identify representative classes of muscle activity patterns. Flight induced changes can then be assessed by comparing postflight trajectories with those obtained prior to flight. Preliminary results suggest that the approach is quite effective in finding groups of gait cycles that show much similarity in EMG activation patterns. For example, Fig. 2 shows the trajectories of the three most similar gait cycles and the two most dissimilar cycles of one astronaut's preflight gait.
Figure 2. Trajectories of the three most similar (top) and most dissimilar gait cycles as identified by the clustering algorithm. Trajectories using three muscles are shown: Bi = Biceps Femoris, Re = Rectus Femoris, and Ti = Tibialis Anterior.
Our approach provides a new way of using EMG to obtain information. As opposed to analyzing time series EMG data with linear system analysis techniques, our multidimensional approach provides an enhanced "landscape" of the neurological control signals used to generate muscle forces during locomotion.
We are currently investigating the properties of our algorithms in more detail. Once that stage of the research is finished, we will process all the available data and devote our attention to developing nonlinear models to relate kinematic data (i.e., position, velocity and acceleration data from leg joints) to EMG data obtained simultaneously (and vice versa). Coupling electrophysiological observations with kinematic data allows us to study the system responsible for gait production from unique viewpoints. Analysis will increase our knowledge of gait production mechanism for future modeling efforts. Once a model is available, the means to intervene can be efficiently developed, i.e., changes in gait patterns by preventive measures or realtime interventions.
Back to Table of Contents
Nicolaos B. Karayiannis, Ph.D., Assistant Professor, Electrical and Computer Engineering, UH
The objective of vector quantization is the representation of a set of random vectors by a set of prototypes. Vector quantization has been proven to be a powerful technique in a variety of applications, including low bit rate compression of digital images. Vector quantization can be performed by learning vector quantization (LVQ) algorithms, whose implementation is associated with a competitive neural network. This network consists of an input layer and an output layer. Each node in the input layer is connected directly to the cells, or units, in the output layer. A weight vector, also referred to as "codebook vector," is assigned to each cell in the output layer.
This project proposed a framework for the development of fuzzy algorithms for learning vector quantization (FALVQ). The development of FALVQ algorithms is based on the minimization of the weighted sum of the squared Euclidean distances between an input vector, which represents a feature vector, and the weight vectors of the network, which represent the prototypes. The distances between each input vector and the prototypes are weighted by a set of membership functions, which regulate the competition between various prototypes for each input and, thus, determine the strength of attraction between each input and the prototypes during the learning process. The design of specific FALVQ algorithms reduces to the selection of membership functions that satisfy certain properties. The general formulation considered in this project provided the basis for the development of the FALVQ 1, FALVQ 2, and FALVQ 3 families of algorithms.
This project also resulted in an alternative methodology for the development of FALVQ algorithms. According to this methodology, the development of a broad variety of FALVQ algorithms can be accomplished by selecting the form of the interference function that determines the effect of the nonwinning prototypes on the attraction between the winning prototype and the input of the network. This methodology provided the basis for the extension of the FALVQ 1, FALVQ 2, and FALVQ 3 families of algorithms.
A clear relationship was established in this project between the form of the membership function used for formulating the LVQ problem and the competition between the prototypes during the learning process. Competition between the winning and nonwinning prototypes can be regulated by simply modifying a single parameter. The selection of this parameter can be facilitated by two quantitative measures that establish a relationship between the formulation that led to FALVQ algorithms and the competition between the prototypes during the learning process. The validity of the proposed competition measures was tested in the limit where the FALVQ algorithms allow only the winning prototype to the updated in order to match the input vector.
The low computational requirements of the proposed FALVQ algorithms make them effective tools in applications involving large numbers of feature vectors of high dimensionality, such as image compression based on vector quantization. The application of the proposed algorithms in codebook design for image compression demonstrated their ability to design high quality vector quantizers for nontrivial tasks. FALVQ algorithms have successfully been used in conjunction with wavelet-based subband image decomposition to compress high-resolution digital mammograms.
Back to Table of Contents
Joan Maier, Ed.D., Assistant Professor, Education, UH
Geography and environmental education in the K-12 curriculum gained renewed emphasis in a graduate teacher-training summer institute for K-12 educators. Sixteen practicing teachers from local, national, and international school districts attended a three-week institute which emphasized instruction in national geography standards enabling spatial and environmental analysis.
Professors of geography provided content on global, physical, and environmental geography. Dr. Joan Maier and two National Geographic teaching consultants assisted the teachers in the development of curricula on global, physical, and environmental geography through the use of experiential problem-solving teaching methods and space-source photography. NASA-contracted Lockheed and Hernandez scientists trained the teachers to interpret space photography.
Sixteen different teaching units were developed by institute participants for the utilization of space photography to teach students global, physical, and environmental geography. Each teaching unit comprised five to ten related lessons on concepts of one general theme or topic. Besides the teaching units, sixteen different warm-up or "sponge" activities that utilize space photography were developed for classroom instruction. Warm-up or "sponge" activities are instructional activities that constitute five to ten minutes of class time. They are used by K-12 teachers to introduce a lesson or to summarize major points of a lesson.
Institute evaluations and follow-up telephone interviews indicated that 15 institute participants plan to use either all or portions of their teaching units during this school year. Sixteen participants indicated that they plan to use several of the warm-up activities with their students throughout this school year. All participants indicated that they plan to use space photography with other lesson materials. Eight participants indicated that they plan to disseminate lessons, warm-up activities, and space photographs to colleagues at their schools, and make them known through regional workshops or state conferences associated with social studies education, geography education, and science education. A second telephone survey during the 1995-1996 school year plans to monitor implementation of the teaching units, warm-up activities, and space photography.
The sixteen Space Photography Teaching Units developed by K-12 educators and grade levels are listed below:
Contents
ISSO -- Institute for Space Systems Operations
1994-1995 Annual Report
|
|
