Manjula B. Waldron

Isolated ASL sign recognition system for deaf persons

The design and evaluation of a two-stage neural network which can recognize isolated ASL signs is given. The input to this network is the hand shape and position data obtained from a DataGlove mounted with a Polhemus sensor. The first level consists of four backpropagation neural networks which can recognize the sign language phonology, namely, the 36 hand shapes, 10 locations, 11 orientations, and 11 hand movements. The recognized phonemes from the beginning, middle, and end of the sign are fed to the second stage which recognizes the actual signs. Both backpropagation and Kohonens self-organizing neural work was used to compare the performance and the expandability of the learned vocabulary. In the current work, six signers with differing hand sizes signed 14 signs which included hand shape, position, and motion fragile and triple robust signs. When a backpropagation network was used for the second stage, the results show that the network was able to recognize these signs with an overall accuracy of 86%. Further, the recognition results were linearly dependent on the size of the finger in relation to the metacarpophalangeal joint and the total length of the hand. When the second stage was a Kohonens self-organizing network, the network could not only recognize the signs with 84% accuracy, but also expand its learned vocabulary through relabeling. >

A time sequence study of a complex mechanical system design

Abstract In the present work the design of a complex mechanical system, namely the leg of a walking machine, is investigated by using a time analysis that uses specially coded boxes to identify the use of important knowledge bases and the decisions made. Based on this study, hypotheses on the generation of sub-goals and the length of the conceptual design process are proposed for future research wotk in mechanical design theory and methods.

Designing instructional formats: the effects of verbal and pictorial components on hearing-impaired students' comprehension of science concepts.

This study compared the effectiveness of four different printed instructional formats on the comprehension of science information by 60 hearing-impaired students aged 12 to 22 with severe to profound hearing losses. The students studied one of four randomly assigned formats describing basic science concepts derived from a sixth-grade science textbook. Comprehension was measured by gain scores calculated from pretest and posttest scores. Results indicated that formats which featured highly pictorial content and simplified English produced significantly higher mean gain scores than formats with less pictorial content and more complex English patterns. Differences were noted also between male and female participants. The study has general implications for the design of instructional materials developed for hearing-impaired learners.

Mechanical Design: Theory and Methodology

From the Publisher: This collection of review articles by active researchers on mechanical design presents an overview of the current state of the art in the field and provides the reader understanding of both the theory and applications of current work in process and product design. Reflecting the almost revolutionary developments in engineering design over the last decade, the contributions to this volume make use of insights from such fields as artificial intelligence, concurrent design, computer science, and psychology. The contributions span the full range of concern of mechanical design: from theory to applications, from processes and products to information flow, from individual to group design, from representation of knowledge domains to general analogic reasoning, and from single-concept design to life-cycle design and issues of quality. Each chapter is by an active researcher who has contributed to the theory or methodology of mechanical design. The chapters are all self-contained and can be read independently from the others; selections of chapters can thus be used in teaching a course on integrated product design.

The Influence of the Designer’s Expertise on the Design Process

Designers bring their prior experience and expertise to the design process whenever they read drawings, draw, or design or when they observe. It is for this reason that one cannot separate the process from the expertise of the designer. What is designed is integrally tied to the designer. The question then arises: “Are there notable differences between experienced and not-so-experienced designers, and if there are, what form do they take?” In this chapter we summarize our previous work on identifying the differences between the approaches of designers with different levels of expertise. These include the differences in recall of drawings, in reasoning about motion, in reasoning about drawings, and in designing. We present the implications of this work on the design process.

Spatiotemporal neural network using axodendritic chemical synapse model

An artificial neural network which can process spatiotemporal signals is developed. In the proposed time-varying neural network, in addition to the weights as used in the classical artificial neural network, the encoding of timing information in the input signal is also considered. The proposed processing unit uses (a) weights as a lumped parameter of synaptic efficacy and attenuation of the postsynaptic potential due to propagation through the dendrite; (b) time delay between the arrival of action potential at the presynaptic site and the generation of action potential at the axon hillock; and (c) decay rate of the postsynaptic potential generated by the previous inputs. These properties are derived from the axodendritic chemical synapse model. Computer simulation of the proposed neural network shows that this network is sensitive to the phase dependencies in the input signals and that it can detect specific sequences in the spatiotemporal input signals.<<ETX>>

An adaptive threshold learning algorithm for classical conditioning

A neuronal model featuring the ability to encode the spatiotemporal relations between input signals is proposed to delineate some of the aspects of classical conditioning. The model uses a spatiotemporal neuron (STEN) and adaptive threshold learning (ATL). During learning, both threshold, and weights are updated as training proceeds. Computer simulations demonstrate that the model exhibits the basic properties of delay and trace conditioning, different ISI effects, blocking, overshadowing, and compound stimulus.

RESPONSE LATENCY MEASUREMENTS TO VISUAL COGNITIVE TASKS BY NORMAL HEARING AND DEAF SUBJECTS

Visual cognitive differences between hearing (N = 16) and deaf (N = 32) high-school and middle-school students were studied. Visual tasks were presented on a microcomputer and response latencies were collected. Significant differences were noted between the deaf and normal groups but not between total communication deaf and oral deaf students. These differences support the hypothesis that deaf students prefer a visual cognitive strategy. Implications for educating the deaf are discussed.

Methods of Studying Mechanical Design

In this chapter we present different methods used for studying mechanical design, including the process of creating designed artifacts. If we wish to understand fully the manner in which successful products are created and design specifications are converted into information for manufacturing and use, then we need to comprehend better the design process that creates them. While successful engineering products have been designed for centuries, understanding the process of this conversion has only begun. This has been largely motivated by a need to create a more efficient process to decrease time to market, requiring better design tools to assist designers in coping with the increasing informational demands placed upon them. The methods of study naturally involve interactions with designers or design teams, be it through case studies, interviews, or observations. In this chapter we survey different methods currently in use to study this process and provide the context in which each method can be used.

Increasing manual sign recognition vocabulary through relabelling

In this paper we present the results of relabelling a self organizing map (SOM) to increase the dynamic manual signs it can recognize. Relabelling exploits the global ordering of self organizing map and abrogates the need for retraining, thereby reducing the computational costs and increasing the recognition ability of the network. This relabelling technique was applied to a dynamic sign recognition system to increase the recognition vocabulary from 10 to 14 signs. The data was collected from a person wearing a DataGlove with a Polhemus sensor and signing the 14 signs. The sampled hand data over the duration of sign was fed to phonemic recognition modules and the collective outputs of these modules were fed to the sign recognition module consisting of a relabelled self organizing network. The results showed that the overall recognition rate of the relabelled network was 84% as compared to 86% for the retrained network. Further, it was found that the dynamic sampling of the signs made the movement phoneme module unnecessary.<<ETX>>