The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #94103

Title:Steerable Ears for Robotic Cat
Date: 1994
Abstract:This dissertation is an account of the possible ways to construct artificial steerable ears for a robotic cat. The problem underlying the specifications of the project is that of sound localization in the real world. Directional hearing is not yet fully understood, and methods by which sound localization might be artificially performed are not yet well known. The first part of this dissertation is thus a comprehensive report of the theory of sound localization. The localisation cues of interaural time and level disparities, and those provided by the spectrum of sounds are explained and discussed. The second part of this thesis deals with two possible methods by which directional hearing can be achieved using the interaural cues. One of these methods if very noise-dependent but easy to implement in hardware, whereas the other is robust but hard to model. They both tackle the correspondence problem by using separate frequency channels, in order to compute the delay in separating the signals. The delay is then transformed into direction. it is argued that the selection between the two methods depends on the measurement of the expected noise in the real world.

[Search These Pages] [DAI Home Page] [Comment]