Non-individualized head-related transfer function (HRTF) devices cannot generate accurate directional sound cues. In particular, listeners often incorrectly perceive a sound cue from the front as coming from the back (referred to as front/back confusion). This study presents a methodology to reduce the occurrence of front/back confusion when listening to sound cues generated using non-individualized HRTF devices. The study investigates and identifies that front/back confusion occurs because of mis-matches between non-individualized HRTFs and the acoustics characteristics of the ears of individual listeners. Using literature concerning the acoustics characteristics of human ears, spectrum of 196 open-copyrighted non-individualized HRTFs have been analyzed and clustered according to thei...[ Read more ]

Non-individualized head-related transfer function (HRTF) devices cannot generate accurate directional sound cues. In particular, listeners often incorrectly perceive a sound cue from the front as coming from the back (referred to as front/back confusion). This study presents a methodology to reduce the occurrence of front/back confusion when listening to sound cues generated using non-individualized HRTF devices. The study investigates and identifies that front/back confusion occurs because of mis-matches between non-individualized HRTFs and the acoustics characteristics of the ears of individual listeners. Using literature concerning the acoustics characteristics of human ears, spectrum of 196 open-copyrighted non-individualized HRTFs have been analyzed and clustered according to their abilities to generated directional sound cues that are coming from the front and from the back. These clusters are further processed to determine six standard directional sound cues for the center-front and center-back directions, respectively. These twelve standard sound cues enable listeners to choose and minimize the possible mis-match that leads to front/back confusion. Three experiments have been conducted and the results indicate that providing the choice of these six standard frontal cues and the six standard backward cues can significantly reduce the sound localization errors and the rates of front/back and back/front confusion. In addition, providing the choice of these twelve cues have been shown to achieve better sound localization performance than using the sound cues generated with the MIT KEMAR non-individualized HRTF set. This finding is important because the MIT KEMAR HRTF set is the most widely-used open-copyrighted HRTF data set. Results reported in this thesis enable engineers to design and develop customizable non-individualized HRTF devices in a cost-effective way. Currently, most digital audio products are manufactured in China but the profit margins for such products are decreasing annually. If industrialists in Hong Kong can acquire the ability to produce customizable virtual surround sound systems using non-individualized HRTF devices at low cost, new innovative products with much greater profit margins can be designed and produced.