Multilingual model-based rehabilitative audiology

听力损失是现代社会面临的最常见的感觉障碍之一，且呈上升趋势。然而，听力设备的诊断和康复方法在各国仍然有限，也没有标准化。缺乏足够的标准化和国际可比的方法，阻碍了改善听力治疗、跨研究进行比较以及更好地了解听力损失对个人在实验室和现实生活中表现的影响的能力。因此，当前项目的主要目标是建立一种统一的、国际兼容的和有理论支持的言语听力学方法，用于听力诊断以及听力装置益处的处方和评估。这将通过将各种语言的基于语音的临床听力诊断与语音清晰度模型的开发和评估联系起来实现。为了确保基于语音的听力诊断在不同语言之间的可比性，矩阵语句测试将应用于与人类的语音识别测量以及基于自动语音识别(ASR)的模型预测(Schädler等人，2015)。MATRIX测试将首先针对不同的语言进行验证，然后包括在用于听力诊断的临床测试组件中。对于客观评估，基于ASR的模型将扩展到不同的语言、双耳听力，并预测具有不同听力状态的听者的语音识别。由于该基于ASR的模型允许预测语音接收阈值，而无需将该模型调谐到任何经验数据，因此将进一步利用该建模概念对基于模型的助听器装配或基于模型的助听器设置的预选的应用。为了支持这种系统的、模型辅助的不同助听器的测试和比较，除了患者自己的助听器外，还将应用Master助听器平台。该平台可以实现不同的算法，并将它们组合到一个模拟助听器中。它将与语音识别测量软件相结合，以可比和可控的方式进行多中心研究。将在几个国际诊所和研究中心实施共同的测量方案，以评估不同助听器的个人益处。该项目将与不同国家的领先诊所和研究中心密切合作，以促进多语言比较，并将有助于在言语测听以及听力设备的处方和评估领域建立国际公认的统一标准。

Hearing loss is one of the most frequent sensory deficits challenging our modern, communication-oriented society with increasing tendency. However, the diagnostics and rehabilitation approaches with hearing devices are still limited and not standardized across countries. The lack of adequate standardized, internationally comparable methods hampers the ability to improve hearing treatment, to draw comparisons across studies, and to better understand the influence of the hearing loss on individual performance in the laboratory and in real life. The main objective of the current project therefore is to establish a unified, internationally compatible and theory-supported approach in speech audiology to be used in hearing diagnostics and the prescription and assessment of hearing device benefit. This will be achieved by linking speech-based clinical hearing diagnostics in various languages and speech intelligibility model development and evaluation. To assure the comparability of speech-based hearing diagnostics across languages, the Matrix sentence test will be applied for speech recognition measurements with humans and for model predictions with Automatic Speech Recognition (ASR)-based model (Schädler et al., 2015). The Matrix test will be first validated for different languages and then included in the clinical test battery used for hearing diagnostics. For objective assessment, the ASR-based model will be extended for the different languages, binaural hearing, and for predicting speech recognition in listeners with different hearing status. Since this ASR-based model allows for predicting the speech reception thresholds without tuning the model to any empirical data, the application of this modelling concept towards model-based hearing aid fitting or model-based preselection of hearing aid settings will be further exploited. In order to support such a systematical, model-assisted testing and comparison of different hearing aids, a Master Hearing Aid Platform will be applied in addition to the patient own hearing aids. This platform enables the implementation of different algorithms and their combination into one simulated hearing aid. It will be combined with software for speech recognition measurements to conduct multicenter-study in a comparable and well-controlled way. A common measurement protocol will be implemented in several international clinics and research centers to assess the individual benefit from different hearing aids. The project will be realized in close cooperation with leading clinics and research centers in different countries to facilitate multilingual comparisons and will contribute to establishing internationally accepted, unified standards in the field of speech audiometry and the prescription and assessment of hearing devices.