Investigating the Feasibility of a Near-Field Binaural Loudspeaker System

Binaural audio techniques have been used since the 1880s to create realistic and convincing virtual audio experiences. However, loudspeaker binaural reproduction generally involves an additional process to artificially increase the separation between the left and right audio channels. This process can affect the quality of the listening experience and complicate the reproduction setup. Therefore, a loudspeaker method which avoids this process would be of significant interest. The approach proposed in this thesis is to passively increase the left-right channel separation through suitable loudspeaker placement. Acoustic simulation enables investigation of more loudspeaker directions and distances for more subjects than is feasible using acoustic measurement. However, simulation requires high-resolution meshes and considerable computational resources. Initial thesis work focused on the development and validation of a suitable 3D mesh model of the human form. A modified version of the mesh was created, comprising only the head and shoulders region. This served to reduce computation time and so enabled simulation of the performance of many loudspeaker positions without reducing the maximum valid simulation frequency. Loudspeaker positions were identified which exhibit left-right channel separation greater than the threshold reported in the literature as required for robust binaural reproduction. To additionally characterise the perceptual impact of loudspeaker placement, the deterioration in interaural binaural cues associated with each loudspeaker pair was determined. Using a conservative model, positions were identified where deterioration was below the estimated threshold of perception for multiple subjects. A small number of loudspeaker pairs at close radial distances met this requirement, but it is likely that these are only a subset of the positions which would perform satisfactorily in a real-world loudspeaker system. This indicates that a simplified binaural loudspeaker reproduction system, capable of satisfactory performance for multiple listeners without adjustment, is a viable possibility.