A. Antonio Arroyo

Classification of Cortical Responses Using Features from Single EEG Records

Our research goal is to develop a new methodology for studying brain function using single, unaveraged EEG records. This investigation has led to a new algorithm for feature extraction for the case of small design (learning) sets. The algorithm has been applied to extract features from unaveraged (single) EEG records, which consist of single evoked responses elicited from human subjects who read textual material presented in the form of propositions. The subjects were instructed to make a binary decision concerning each proposition. This gave two possible data classes. We selected features from the evoked event-related potentials (ERPs), and designed a classifier to assign the ERPs for each proposition to one of the two possible classes.

Team Gator Nation's Autonomous Vehicle Development for the 2007 DARPA Urban Challenge

This paper describes the system design developed for Team Gator Nation’s submission to the 2007 DARPA Urban Challenge. A hybrid Toyota Highlander has been automated and instrumented with pose estimation (GPS and inertial) and object detection (vision and ladar) sensors. The control architecture consists of four primary elements, i.e. Planning Element, Perception Element, Intelligence Element, and Control Element. The Intelligence Element implements the Adaptive Planning Framework developed by researchers at the University of Florida. This framework provides a means for situation assessment, behavior mode evaluation, and behavior selection and execution. The architecture is implemented on a system distributed over ten dual-core computers that intercommunicate via the Joint Architecture for Unmanned Systems (JAUS) version 3.2 protocol. This work’s primary contribution addresses the technical challenges of (a) the reconciliation of differences in estimated global pose, a priori data, and sensed information, (b) the determination of the appropriate behavior mode, and (c) the smooth transition of vehicle control between behavior modes. The processes that perform these tasks as well as the other necessary processes that perform perception, data integration, planning, and control are described in detail together with their design rationale. Finally, testing results accomplished to date are presented.

Acoustic variability in vibrato and its perceptual significance

Summary Historically, studies of vocal vibrato have concentrated on pulse rate as being a primary factor in determining whether a given vocal movement is a good or bad vibrato or a tremolo or wobble. More recently, investigators have been studying the extent of frequency variation and amplitude variation around their respective means in order to determine their influence on the perception of vibrato. The present study is an additional attempt to understand the three parameters comprising vibrato, their interrelationship, and their relationship to perception. Samples of sustained sung tones were obtained primarily from recordings. The samples were digitized using a 16-bit A/D converter at a sampling frequency of 10 kHz. Each digitized sample was converted to a useful format for marking purposes in order to derive information on vibrato pulse rate, the mean frequency of the tone, the semitone deviation around the mean, percent frequency deviation and percent amplitude variation around the mean amplitude. Data presentation utilizes representative samples of good vibrato, tremolo and wobble and describes differences in waveforms which may impact on perception.

A knowledge-based clinical toxicology consultant for diagnosing multiple exposures

OBJECTIVE This paper presents continued research toward the development of a knowledge-based system for the diagnosis of human toxic exposures. In particular, this research focuses on the challenging task of diagnosing exposures to multiple toxins. Although only 10% of toxic exposures in the United States involve multiple toxins, multiple exposures account for more than half of all toxin-related fatalities. Using simple medical mathematics, we seek to produce a practical decision support system capable of supplying useful information to aid in the diagnosis of complex cases involving multiple unknown substances. METHODS The system is automatically trained using data mining techniques to extract prior probabilities and likelihood ratios from a database managed by the Florida Poison Information Center (FPIC). When supplied with observed clinical effects, the system produces a ranked list of the most plausible toxic exposures. During testing, the system diagnosed toxins at three levels: identifying the substance, identifying the toxins major and minor categories, and identifying the toxins major category alone. To enable comparison between these three levels, accuracy was calculated as the percentage of exposures correctly identified in top 10% of trained diagnoses. RESULTS System evaluation utilized a dataset of 8901 multiple exposure cases and 37,617 single exposure cases. Initial system testing using only multiple exposure cases yielded poor results, with diagnosis accuracies ranging from 18.5% to 50.1%. Further investigation revealed that the systems inability to diagnose multiple disorders resulted from insufficient data and that the clinical effects observed in multiple exposures are dominated by a single substance. Including single exposures when training, the system achieved accuracies as high as 83.5% when diagnosing the primary contributors in multiple exposure cases by substance, 86.9% when diagnosing by major and minor categories, and 79.9% when diagnosing by major category alone. CONCLUSIONS Although the system failed to completely diagnose exposures to multiple toxins, the ability to identify the primary contributor in such cases may prove valuable in aiding medical personnel as they seek to diagnose and treat patients. As time passes and more cases are added to the FPIC database, we believe system accuracy will continue to improve, producing a viable decision support system for clinical toxicology.

Multichannel, Single Tral Event Related Potential Classification

We report on the effect of electrode placement and number of electrodes on the classification of single trial event related potentials (ERPs). The subjects read propositions relating fictitious people and their occupations while ERPs were recorded. The subjects decided if the proposition was correct or incorrect and responded as per instructions. The single trial, multichannel ERP data were classified using various methods, e. g., hold-out, leave-one-out, resubstitution. Several other factors were examined to determine their effect on ERP classification, including taking a majority vote among channels, using the single best channel, and averaging the data across channels for a single ERP. The results from other experiments are compared to those presented here.

A mathematical model of singers’ vibrato based on waveform analysis

There is controversy regarding the best model to synthesize vibrato. Some authors include random components in their models while others suggest a deterministic model. Also, the vibrato waveform has not been studied in detail, that is, no measures have been made as to symmetry of the waveform, and parameters different from wave frequency and amplitude have not been studied. Therefore, research was undertaken to thoroughly analyze the vibrato waveform from selected singers in order to obtain a model based on the vibrato parameters. Seven premier‐level singers were selected and three samples per singer were used. These samples were analyzed through the MMSV (Mathematical Model of Singers’ Vibrato) software, which was specifically designed for vibrato analysis. The results obtained for all the samples were analyzed and compared, and a model fitting all of them was proposed. This model was compared against a sinusoidal model and the real wave for two different samples by the use of an error measure and visually, and it was verified that the model reproduced the variations of the real vibrato wave and significantly reduced the error measure of the sinusoidal model. The results show that a deterministic model fits all the samples under study.

An expert system for academic counseling

In recent years, high enrollment has greatly increased the counseling loads of academic advisors at American colleges and universities. Expert systems can assist academic counselors by solving a relatively easy class of problems which deal with the most encountered cases. In this paper, a prototype expert system, CLASS COUNSELOR, is presented. CLASS COUNSELOR recommends a set of courses after an interactive session with the student. It handles the upper division courses of the undergraduate program in the Department of Electrical Engineering, University of Florida. The program runs on IBM-AT personal computers and compatibles.

Reading up books: Artificial intelligence and expert systems: Academic guidelines and suggested reading in the field of artificial intelligence and its industrial arm — Expert systems

“Intelligence … is the faculty of making artificial objects, especially tools, to make tools.” (Henri Bergson, 1859–1941.) What, exactly, is meant by the term “artificial intelligence” (AI)? Dr. Patrick H. Winston, Professor of Computer Science and Director of the Artificial Intelligence Laboratory at the Massachusetts Institute of Technology, defines AI as “the study of ideas which enable computers to do the things that make people seem intelligent. The central goals of AI are to make computers more useful and to understand the principles which make intelligence possible.”

Machine intelligence grounded in reality

The design, realization, and application of intelligent, autonomous, sensor-driven, behavior-based robotic agents are discussed. The authors have identified four philosophical goals for machine intelligence grounded in reality to flourish: integration, real-world issues, interdisciplinary teamwork, and critical thinking. Building a robot requires that the designer integrates control in electronic and mechanical systems and produces a working device, confronts the user with interactions, between different subsystems, and gives the designer the opportunity to trade off between the different subsystems in constructing an autonomous agent. The environment also encourages biomedical engineers to confront the issues involved in getting a physical agent to operate reliably in a realistic environment by giving them the opportunity to build their own animal, providing a unique perspective on the many problems that nervous systems actually solve.

Creation and analysis of a scenario based universal sensory driver layer with real-time fault tolerant properties

Sensor fusion and sensor integration is becoming an increasingly popular approach in dealing with complex sensor systems in autonomous mobile robots (AMR). However, the procedure for the sensor integration and sensor fusion is a non-trivial process. This paper presents a scenario based approach to sensor fusion based on the autonomous evolution of sensory and actuator driver layers through environmental constraints (AEDEC) [T.A Choi, 2002]. Using the scenario based approach, the programmers work of creating a sensory driver will be eliminated by having the AMR learn the driver on its own. In the process of creating each scenario, sensor fusion is automatically implemented. If sensors change or even if the sensor configuration changes, the driver can be updated by having the AMR relearn the driver over again. Due to the tabular structure of the scenario based sensory drivers, malfunctioning sensors can not only be detected, but the driver can automatically adapt to the malfunctioning sensor in real-time. Furthermore, different AMRs trained using AEDEC architecture will have similar interpretations of its environment. This is guaranteed by having the AMR learn the driver in the same highly structured training environment. The behavioral coding is simplified by eliminating any reference to hardware dependent parameters. Finally, the level of abstraction and the consistency of the highly structured environment allows for coding portability.