Back to list

Sound-Source Localisation with Frequency-Warped Head-Related Transfer Functions

Authors Walder, T.
Year 2010
Thesis Type Diploma thesis
Topic Psychoacoustics
Keywords technical acoustics, audio reproduction, acoustics, 3D sound, signal processing
Abstract Electrical stimulation of the cochlear nerve with cochlear implants (CIs) enables deafened people to understand speech and therefore has been shown to successfully substitute basic functions of the inner ear. However, current CI systems can not recover all functions of the normal auditory system. Localisation within the sagittal planes still remains difficult with current CI systems. In normal hearing, this localisation is based on analysing spectral colouration in the high frequency region of the incoming sound, caused by the pinna. In CI systems the implanted electrodes do not cover the relevant frequency region. A possible approach to transmit this spatial information is to warp it into the lower frequency band. This leads to the basic question whether it is possible to relearn localisation by using frequency-warped spatial information. In this project the effect of linear frequency warping of spectral spatial information on localisation of sound sources is studied. To exclude individual factors commonly occurring amongst CI users, listeners with normal hearing participated in this project. Audio stimuli were presented via virtual acoustics using individual head-related transfer-functions (HRTFs) and visual stimuli via a head mounted display. The high frequency band (2.8 - 16kHz) was warped into the frequency band of current CI systems available for localisation (2.8 - 8.5kHz). The low frequency band (0.3 - 2.8kHz), relevant for speech intelligibility, was not modified. 15 subjects with normal hearing, subdivided into a test group and a control group, completed a 21 day audio-visual training in units of 2 hours per day. Whilst the test group practised with frequency-warped HRTFs, the control group practised with merely bandlimited HRTFs (0.3 - 8.5kHz). The decrease of front/back confusions was significantly larger for the test group than for the control group. Nevertheless, the residual polar localisation errors of the test group at the end of the training were larger compared to those measured with original HRTFs. Surprisingly, also the control group showed a large training effect. The localisation error at the end of the training was not significantly different from the localisation error with original HRTFs. The strong auditory plasticity in sagittal localisation opens new ways to improve spatial hearing with CIs and other hearing devices with limited bandwidth.
Supervisors Höldrich, R., Majdak, P., Laback, B.