Optimal Source Distribution for Two-Listener Virtual Sound Imaging

Virtual sound images can be perceived with remarkable clarity when carefully designed signal processing is combined with appropriate arrangements of loudspeakers. Most successful methods to date have made use of crosstalk cancellation at the ears of a single listener to deliver acoustic signals that are a good replica of those that would be produced by a specified virtual source. This paper further investigates the use of crosstalk cancellation for two listeners and produces guidelines for the successful implementation of such systems. The basis of the work presented is a four-source two-listener model that makes the assumption of free field radiation from monopole sources and neglects scattering from the listener’s heads. This simple model is used to examine the conditioning of the matrix of transfer functions relating the source outputs to the listener ear signals. Some new results are derived that relate the matrix conditioning to the differences in path length between each of the four sources and the ears of the two listeners. The relationship between these path length differences and the acoustic wavelength determines the minimum condition number. These results enable the design of loudspeaker arrays that produce highly effective crosstalk cancellation for both listeners over a wide frequency range. Experimental results are presented that demonstrate the effectiveness of this design process.