Independent Component Analysis

Abstract

ABSTRACT The application of auditory steady–state responses(ASSRs) is an objective and frequency specific techniqueto assess reliable hearing thresholds at audiometric fre-quencies. Unfortunately, the duration of ASSR measure-ment can be long, which is unpractical for wide scale clin-ical application. Therefore, we propose independent com-ponent analysis (ICA) as a tool to improve the ASSR de-tection in recorded multi–channel electroencephalogram(EEG) data obtained at high intensities. We conclude thatICA is able to reduce measurement duration significantly.For a multi–channel implementation, near–optimal perfor-mance is obtained with five–channel recordings. A pro-posal for a generative model is suggested. This modelmay provide the ability to explain some observed char-acteristics of the ICA–ASSR problem.Keywords: independent component analysis, audi-tory steady-state response, electroencephalogram, multi-channel. 1 INTRODUCTION Children with hearing problems have to be diagnosed andsupported by hearing aids as early as possible [10]. There-fore, the hearing capability of 96.5 % of all newbornsin Flanders is effectively screened by the Flemish childcare organisation Kind & Gezin (‘Child & Family’) since1998. For the fitting of a hearing aid and for general di-agnostic purposes, it is however necessary to quantify thehearing thresholds of these children in an objective way.In the late eighties, several techniques using AuditorySteady–State Responses (ASSRs) were developed to per-form objective and frequency specific threshold measure-ments. ASSRs are faint evoked electrical responses of thebrain [12]. These responses can be elicited by amplitude–and/or frequency–modulated (AM/FM) pure tones. If a