P1: FRP 3053 NECO.cls June 17, 2005 0:15 LETTER Communicated by Shihab Shamma A Novel Model-Based Hearing Compensation Design Using a Gradient-Free Optimization Method Zhe Chen zhechen@soma.ece.mcmaster.ca Department of Electrical and Computer Engineering, McMaster University Hamilton, Ontario L85 4k1, Canada Suzanna Becker becker@mcmaster.ca Department of Psychology, McMaster University Hamilton, Ontario L85 4k1, Canada Jeff Bondy jeff@soma.ece.mcmaster.ca Department of Electrical and Computer Engineering, McMaster University Hamilton, Ontario L85 4k1, Canada Ian C. Bruce ibruce@ieee.org Department of Electrical and Computer Engineering, McMaster University Hamilton, Ontario L85 4k1, Canada Simon Haykin haykin@mcmaster.ca Department of Electrical and Computer Engineering, McMaster University Hamilton, Ontario L85 4k1, Canada We propose a novel model-based hearing compensation strategy and gradient-free optimization procedure for a learning-based hearing aid design. Motivated by physiological data and normal and impaired au- ditory nerve models, a hearing compensation strategy is cast as a neural coding problem, and a Neurocompensator is designed to compensate for the hearing loss and enhance the speech. With the goal of learning the Neurocompensator parameters, we use a gradient-free optimization pro- cedure, an improved version of the ALOPEX that we have developed (Haykin, Chen, & Becker, 2004), to learn the unknown parameters of the Neurocompensator. We present our methodology, learning procedure, and experimental results in detail; discussion is also given regarding the unsupervised learning and optimization methods. Neural Computation 17, 1–24 (2005) © 2005 Massachusetts Institute of Technology