A CLUSTERING METHOD FOR SOUND LOCALIZATION TRANSFER-FUNCTIONS

A new algorithm is proposed which prepares a set of sound localization transfer functions (SLTFs) for a sound localization system with binau ral earphones. A listener in a teleconference system can choose the be st SLTFs from the set prepared in advance. This avoids the trouble of measuring SLTFs individually and maintains good perceptual localizatio n performance. This algorithm reduces the vector order of the individu al SLTFs, for example, from 512 to 16, using cepstrum parameters. The SLTF vectors are classified into several clusters using the Linde, Buz o, and Gray method, and the SLTF nearest each cluster's centroid is ch osen as the typical SLTF in the cluster. The subjective performance of sound localization is shown in which the proposed algorithm selects t ypical SLTFs from among 94 measured SLTFs.