Auditory Detection and Sound Localization for Computer-Generated Individual Combatants

摘要

Soldiers rely predominantly on vision to detect targets yet other senses may cue their sense of sight. Contrarily most army combat simulations employ only visual cues. The focus of this thesis is to enhance combat simulations by providing a method by which computer-generated entities can detect and locate objects via a phenomenon known as 'sound localization.' The Auditory Detection Program is used to represent a human's hearing and data from a sound localization experiment are analyzed to determine how to best represent the event in which an individual hears a sound and then estimates the location of the sound's source. The resulting algorithms are coded into the Army's combat simulation COMBAT(XXI) and the face-validation method is used to determine if the algorithms enhance the realism of the simulation. The data analysis consists of Shapiro-Wilks Tests for Normaility Friedman's Tests for Randomized Block Experiment, and Wilcoxon Rank-Sum Tests using the Bonferroni Correction. Implementing this model in COMBAT(XXI) improves the simulation by making it more realistic.