Human-liked Auditory Capability for Intelligent Virtual Agents
(Room 320A)
04 Dec 19
10:30 AM
-
11:00 AM
Tracks:
Full Schedule, Wednesday Schedule
Intelligent Virtual Agents (IVAs) are important components in simulated real-world environments. Usage of IVAs in training is mainly for task collaboration where virtual agents interact with each other or with human users. Last year’s paper “Human-liked Auditory Detection Capability for Intelligent Virtual Agents” (I/ITSEC 2018) presented an auditory perceptual model that can be used to predict the capability to detect sound cues in noisy environments. Besides the capability to detect sound cues, the human hearing system has the ability to identify the direction from which the sound is coming, estimate the distance of the sound source and eventually assess the characteristics of the physical surrounding environment affecting sound propagation (Auditory Spatial Perception). Auditory localization represents the most critical element of the auditory spatial perception for human effectiveness and safety.
An IVA auditory perception model without a capability to localize sound sources is incomplete. To complement the previously completed auditory capability modeling of an IVA from last year’s paper, a perceptual model will be added to simulate the capability of an IVA to localize sound sources. First, the paper will provide the foundation of this perceptual model, which is based on the Duplex theory of the human hearing system - Interaural Time Difference (ITD) and Interaural Level Difference (ILD). Then, it will explain how this model was integrated with the IVA model to simulate the sound localization capability. Finally, the paper will present the simulation results and assess the effectiveness of this auditory perceptual model when used with the simulation of an IVA. Preliminary analysis of simulation results indicated that Duplex Theory is suitable to simulate the capability to localize sound sources, especially when the Signal-to-Noise (S/N) ratio is favorable to the sound localization task, e.g. S/N is equal or greater than 12 dB SPL.