Back to list

Binaural Sound Reproduction via Distributed Loudspeaker Systems

Authors Teschl, M.
Year 2000
Thesis Type Diploma thesis
Topic Spatial Audio
Keywords audio reproduction
Abstract The basic principle of binaural sound reproduction technique is to reconstruct the same sound pressures at a listener's eardrums that would have caused there by a real sound source to be simulated. Consequently, the listener cannot distinguish between the real sound source and the generated virtual sound source. If a pair of loudspeakers is used, the appropriate ear signals are delivered to the listener by inverting the transmission paths between the two loudspeakers and the two ears. This process, known as „crosstalk cancellation“, can be considered as an inversion of a [2 × 2] matrix of transfer functions. Previous work undertaken in this area was concentrated on the use of a conventional stereo set-up where the loudspeakers span an angle of 60° as seen from the listener. As opposed to a stereo set-up, by using two closely spaced loudspeakers, the performance had proven more robust with respect to misalignment or movement of the listener's head. However, one disadvantage of this approach is the source strengths required for the crosstalk cancellation at low frequencies. In terms of matrix algebra, the crosstalk cancellation problem is said to be „ill-conditioned“ at these frequencies. Based on a free-field model of the problem, it can be shown that ill-conditioning depends on frequency and the loudspeaker span of the system, respectively. For instance, for a smaller span the system inversion is ill-conditioned at low frequencies, whereas for larger source spans the conditioning is worse at higher frequencies. This connection resulted in the idea to vary the source span as a function of frequency in order to maintain the best possible conditioning over the whole frequency range. A practical solution of this new approach is to use multiple pairs of loudspeakers for each frequency range with corresponding source spans in order to eventually cover the whole audible frequency range. This diploma thesis will discuss the potential for such an approach. Theory, practical implementation, and testing of such systems will be described in detail. Many sound localization experiments were conducted in order to subjectively validate the system's performance. Results show a significant improvement, in particular with respect to azimuth localization for virtual images well to the sides.
Supervisors Höldrich, R., Takeuchi, T., Nelson, P.