Dual-channel equalization and cross talk cancellation for audio applications
Canfield, Gregory H.
MetadataShow full item record
Three-dimensional sound effects depend on psychoacoustic spectral and phase cues being present in a reproduced signal. Cross-talk cancellation and dual-channel equalization using an adaptive filter network have the potential of greatly improving the three-dimensional perception of sound reproduced over two-channel loudspeaker systems. This thesis studies issues related to dual-channel equalization and cross-talk cancellation. First, some general psychoacoustic principles that lead to an accurate perception of the location of a sound source are discussed. Next, traditional methods of cross-talk cancellation and dual-channel equalization are introduced, and problems that correlated input signals pose for these algorithms are demonstrated. The dual-stage algorithm is developed to improve performance in the presence of these types of signals. The dualstage algorithm first models the direct path with a modified dual-channel modeling algorithm that uses a low level of additive uncorrelated, and then forms the inverse of the modeled path. Problems associated with inverting systems with nonminimum phase are discussed. Experimental results are presented from equalizing a signal and playing it through two loudspeakers in a real room. Both analytical and informal perceptual observations of the equalized signal are presented and discussed. This thesis addresses issues unresolved in previously published research, permitting movement toward a dual-channel equalization and cross-talk cancellation system that can function in a real-world environment.