Start Date: | 5/23/2014 | Start Time: | 3:30 PM |
End Date: | 5/23/2014 | End Time: | 5:00 PM |
|
Event Description Title: Assessing Performance of Data Fusion Algorithms using Human Response Models Advisor: Dr. Moshe Kam Date: Friday, May 23, 2014 Time: 3:30 p.m. Location: ECE Conference Room 302, 3rd Floor, Bossone Research Enterprise Center
Abstract
There is increasing interest in designing data fusion systems which make use of human opinions (i.e., ”soft” data) alongside readings from electrical, optical, and mechanical sensors (i.e., ”hard” data). One of the major challenges in the development of these systems is to determine accurately and flexibly the impact of human decisions and confidence assessments on the performance of a fusion operator. Examples and counterexamples have been used to illustrate specific performance aspects of a fusion operator in different scenarios. However these examples do not provide a systematic means for calculating performance statistics, nor are they easy to manipulate by changing parameters in the observed environment. Data sets of human responses developed through human testing have been used to some extent to estimate fusion system performance statistics, however the results obtained in this manner are often hard to generalize and difficult to tune up due to the experimental and administrative limitations imposed on human testing.
Models of human decision making and confidence assessment from cognitive psychology offer a unique opportunity to assess the performance of systems which make use of human opinions. Such models can be used for assessing the performance of hard/soft fusion systems and to make credit assignments to the multiple sources of data that lead to the final estimate/decision of the fusion architecture. Some validation with human decision makers may still be needed, but a lot of design, parameter selection, tuning, as well as stability and convergence analyses can be significantly accelerated and be made more efficient than if one was to rely solely on human testing or on researcher intuition.
The main contribution of this thesis is the development of methods for ascertaining the performance of various soft and hard/soft fusion techniques using models of human responses available in the cognitive psychology literature. We also present methodologies for extending the human response simulation techniques to multihypothesis tasks, and for simulating imprecise responses. |
|
Location: ECE Conference Room 302, 3rd Floor, Bossone Research Enterprise Center |
Audience: Current StudentsFacultyStaffGraduate Students |
|