Computer system for development of a neural compensator based hearing aid

麦克马斯特大学心理学、神经科学与行为学系苏·贝克尔教授领导的一个研究小组与教授们合作。欧洲经济学院的Haykin和Bruce，以及研究生Jeff Bondy和喆Chen，开发了一种新的听力补偿算法，称为NeuroCompensator。基于一个全新的想法，NeuroCompensator的设计目标是使用机器学习方法和正常和受损耳朵的精确生物模型来创造一种助听器，它将恢复听力受损者听觉系统中接近正常的神经元活动。神经补偿器和其他算法已经被麦克马斯特团队申请了专利。美国神经补偿器专利于2006年12月12日发布，专利号为：7149320年。这项专利于2006年8月3日被指定给麦克马斯特大学，并授权给CAYCe医疗公司。CAYCe医疗有限公司拥有该技术商业化的独家全球许可。将该技术转化为商业产品需要以下步骤：1)与标准助听器信号预处理步骤集成，如响度压缩；2)开发广泛的语音和其他声音样本训练集；3)评估是否需要一个或多个神经补偿器来拟合训练集；4)针对不同的听力损失情况，训练几种不同版本的神经补偿器；5)在低功耗助听器微处理器上的实现。听力学家将在一张CD上收到最后一组预先训练好的神经补偿器参数，以及训练软件和音频训练样本。为一个人配戴基于NeuroCompensator的助听器的过程只需要选择与人的听图最匹配的NeuroCompensator，可选地运行软件来微调参数，并将最终的NeuroCompensator参数加载到助听器上。在OCE的市场准备补助金的资助下，Becker现在正在与CAYCe医疗公司合作，完成神经补偿器的研究和开发，并将这项技术商业化。OCE提供工资支持，同时要求NSERC RTI拨款购买这项工作所需的设备。

A McMaster team of researchers led by Prof. Sue Becker, Dept. of Psychology Neuroscience & Behavior, in collaboration with Profs. Haykin and Bruce in ECE, and graduate students Jeff Bondy and Zhe Chen, developed a novel algorithm for hearing compensation called the NeuroCompensator. Based on an entirely new idea, the design goal of the NeuroCompensator is to use machine learning methods and accurate biological models of the normal and impaired ear to create a hearing aid that will restore near-to-normal neuronal activity in the auditory system of the hearing impaired person. The Neurocompensator and other algorithms have been patented by the McMaster team. The US NeuroCompensator patent was issued on Dec. 12, 2006 under patent no. 7,149,320. This patent is assigned to McMaster University and licensed to CAYCe Medical Inc. on August 3rd 2006. CAYCe Medical Inc. holds an exclusive worldwide license to commercialize this technology.  Translation of the technology into a commercial product requires the following steps: 1) integration with standard hearing aid signal preprocessing steps such as loudness compression; 2) developing a broad training set of speech and other sound samples; 3) evaluate whether one or more NeuroCompensators are required to fit the training set; 4) train several different versions of the NeuroCompensator on different hearing loss profiles; and 5) implementation on a low-power hearing-aid microprocessor. An audiologist would receive the final set of pre-trained NeuroCompensator parameters on a CD, together with the training software and audio training samples. The process of fitting an individual for a NeuroCompensator-based hearing aid would simply involve selecting the NeuroCompensator most closely matched to the person's audiogram, optionally running the software to fine-tune the parameters, and loading the final NeuroCompensator parameters onto the hearing aid. Funded by a market readiness grant from the OCE, Becker is now working with CAYCe Medical to complete the research and development of the NeuroCompensator and commercialize this technology. OCE is providing the salary support, while the NSERC RTI grant is requested to purchase the necessary equipment for this work.