IIR modeling of interpositional transfer functions with a genetic algorithm aided by an adaptive filter for the purpose of altering free-field sound localization

The psychoacoustic process of sound localization is a system of complex analysis. Scientists have found evidence that both binaural and monaural cues are responsible for determining the angles of elevation and azimuth which represent a sound source. Engineers have successfully used these cues to build mathematical localization systems. Research has indicated that spectral cues play an important role in 3-d localization. Therefore, it seems conceivable to design a filtering system which can alter the localization of a sound source, either for correctional purposes or listener preference. Such filters, known as Interpositional Transfer Functions, can be formed from division in the z-domain of Head-related Transfer Functions. HRTF’s represent the free-field response of the human body to sound processed by the ears. In filtering applications, the use of IIR filters is often favored over that of FIR filters due to their preservation of resolution while minimizing the number of required coefficients. Several methods exist for creating IIR filters from their representative FIR counterparts. For complicated filters, genetic algorithms (GAs) have proven effective. The research summarized in this thesis combines the past efforts of researchers in the fields of sound localization, genetic algorithms, and adaptive filtering. It represents the initial stage in the development of a practical system for future hardware implementation which uses a genetic algorithm as a driving engine. Under ideal conditions, an IIR filter design system has been demonstrated to successfully model several IPTF pairs which alter sound localization when applied to non-minimum phase HRTF’s obtained from free-field measurement