Modelling speech communication in noise

本研究的目标是开发定量的方法和模型来预测语音清晰度在现实的情况下，噪声，同时考虑到生产和感知方面的沟通。一种基于噪声听力测试（HINT）的方法，最近被用来建立渔业和海洋部（加拿大）海上和陆地人员的听力标准，将作为这项研究的基本建模框架。需要进行更多的研究，以进一步验证和扩展，使其更广泛地使用。具体的研究项目将集中在以下领域：（1）作为环境噪声特性的函数的说话者的声音输出（电平，语音频谱）的建模，（2）提高，大声和重复语音的可懂度，以及（3）将室内声学模型（语音可懂度指数，语音传输指数）集成到仿真框架中。实验方法将包括主观测试，语音估计工具的开发，以及统计和声学建模方法的使用。在真实的生活中有许多应用，对于这些应用，这种定量方法是重要的。例如，在几种职业环境中，语音通信在嘈杂的环境中以及在困难或危险的情况下（例如，飞机飞行员、警察和其他执法人员、军事人员等）执行。在这些具有挑战性的环境中工作的个人可能会对自己和他人构成安全风险，如果由于错误的沟通而无法正确执行任务。考虑到工人的感知和认知能力以及房间或噪声环境的声学方面的定量方法对于建立人员的听力标准是必要的。同样，像本研究中将要开发的模型在听力学诊所评估助听器或其他康复策略在真实的听力情况下的益处将是重要的。

The goal of this research is to develop quantitative methods and models to predict speech intelligibility in noise under realistic scenarios, taken into account both the production and perceptual aspects of the communication. A method based on the Hearing in Noise Test (HINT) and recently used to establish hearing standards for the seagoing and land-based personnel at the Department of Fisheries and Oceans (Canada) will serve as the basic modelling framework for this research. Additional research is necessary for further validation and expansion for more general use. Specific research projects will focus on the following areas: (1) the modeling of the vocal output (level, speech spectrum) of the talker as a function of the environmental noise characteristics, (2) the intelligibility of raised, shouted and repeated speech, and the (3) integration of room acoustic models (Speech Intelligibility Index, Speech Transmission Index) into the simulation framework. The experimental approach will include subjective testing, the development of speech estimation tools, and the use of statistical and acoustical modelling methods. There are many applications in real life for which such quantitative methods are important. In several occupational settings, for example, speech communication is performed in noisy environments and under difficult or dangerous situations (e.g. airplane pilots, police and other law enforcement officers, military personnel, etc.). Individuals working in these challenging environments may pose a safety risk to themselves and to others if the tasks are not performed properly due to miscommunications. Quantitative methods that take into account the perceptual and cognitive abilities of the workers together with the acoustical aspects of the room or noise environments are necessary to establish hearing standards for the personnel. Likewise, models like the ones to be developed in this research will be important in the audiology clinic to assess the benefits of hearing aids or other rehabilitation strategies in real listening situations.