Binaural hearing is a complex process that is significantly influenced by physical attributes of the listener. These dependencies on the shape of the torso, head, but also the resonance properties of the pinna can be summarized in a digital filter, the so-called head-related transfer function (HRTF). Such HRTFs are not only widely used in the entertainment industry, for example for virtual reality, but also in medical applications like hearing aid calibration. However, the fact that HRTFs must be highly individual for a realistic listening experience is a major obstacle to their usage. Especially deviations in the dimensions of the head and pinna have a significant impact on the subject’s orientation ability.

Individualized HRTFs are an alternative to individual filters, which can only be recorded requiring considerable time and computational effort, as well as special equipment such as an anechoic chamber. An established approximation procedure of such individual transfer functions is the principal component analysis (PCA), a statistical method which is used for reconstructing HRTFs depending on the listener’s anthropometric dimensions. Another individualization procedure examined in this work is frequency scaling.

However, those simulated filters do not represent the original HRTFs flawlessly. Due to their high dimensionality, HRTFs comparison is not an intuitive task and comprehensive visualization is challenging. To enable a meaningful comparison, so called distance measures are required. Although numerous distance measures have been developed, there have been few comparative studies on their attributes.

In this thesis a better understanding of the behavior of distance measures shall be achieved. Various distance measures are presented and evaluated. Subsequently, simulated HRTFs are varied in a controlled manner to display the behavior of the different distance measures. These insights are finally applied to several HRTF measurements to compare measurement setups in a purposeful way.

Melden Sie sich Hier an um die Einladung zu dem Vortrag per E-Mail zu erhalten.
Register here to receive the invitation to this talk via E-Mail.

Zoom-Meeting-ID: 912 7408 8155
Passwort: 130862