A Wiener filter approach to the binaural reproduction of stereo sound.

This paper considers authentic binaural reproduction with loudspeakers using cross-talk cancellation. A systematic time domain deconvolution method is presented using both stochastic and deterministic Wiener filters. This method has advantages in that the Wiener filter is inherently causal and stable so that no additional stabilization process, as generally required in the frequency domain methods, is needed. Errors involved in the authentic reproduction are classified into modeling, inverse, and repeatability errors. In addition, the reproduction error is used as the performance measure of a reproduction system. Finally, a series of monaural and binaural reproduction tests were conducted in an anechoic chamber using a PC based reproduction system with a soundcard by changing the length of the inverse filter used. Test results demonstrated that the performance was improved in proportion to the length of the filter used. The reproduction performance error and its test procedure presented here can be used for objectively assessing the performance of a subjective auditory perception test system.