Timing of turn taking in conversation to assess listening effort and evaluate hearing-aid processing

语言是人类交流最重要的部分之一，也是社会互动的关键要素。然而，对许多人来说，口语的使用是一个挑战。十分之一的加拿大人有言语、语言或听力障碍。这个问题随着年龄的增长而加剧，60到69岁的人中有40%有某种类型的沟通障碍，70到79岁的人中有65%有某种类型的沟通障碍。传统上，听力受损（HI）听众在嘈杂环境中交流时遇到的困难归因于与正常听力（NH）听众相比，言语清晰度降低。因此，该领域的重点是涉及被动听力测试的实验范式。然而，谈话是一个复杂的现象，不仅涉及言语的产生和感知，而且涉及谈话者对彼此和环境的互动和适应。拟议的研究计划旨在通过对互动对话的关键要素获得新的见解来解决这一挑战。本研究将调查谈话中轮流的时间。为了实现日常对话中典型的快速轮换，说话者需要同时做三件事：(1)理解他们的伙伴正在说的话，(2)计划一个演讲回应，(3)预测他们的伙伴什么时候会结束他们的轮换。因此，当听力变得更加困难时（例如，由于噪音或听力丧失），说话者可能无法保持在安静的谈话中典型的轮流时间。本研究有四个目标：(1)表征噪声和听力损失对会话中回合结束时间预测的影响；(2)描述助听器可以用来估计听者在对话中遇到的困难的声学标记；(3)制定基于实验室的对话听力努力测量方法；(4)建立基于会话测量的助听器加工评价范式。在短期内，这里的结果将导致更好地评估新型助听器功能和信号处理算法的新方法。此外，了解正常听力者和听力受损者的轮流行为有何不同，可以用来开发用于咨询的新沟通策略。通过将研究成果与学生培训的发展相结合，拟议的研究计划将支持下一代助听器的开发，显着改善老龄化人口的生活质量，提高退休工人的生产力，并使他人更容易与十分之一的加拿大人沟通言语，语言或听力障碍。最后，提高我们对对话中交互行为的基本理解也将影响其他领域，例如开发更自然的人机语音交互系统。

Speech is one of the most important parts of human communication and a key element of social interaction. However, for many people, the use of spoken language is a challenge. One in 10 Canadians has a speech, language or hearing impairment. This problem is exacerbated with age with 40% of those between 60 and 69 and 65% of those between 70 and 79 having some type of communication disorder. Traditionally, the difficulty hearing-impaired (HI) listeners experience when communicating in noisy environments has been attributed to reduced speech intelligibility compared to normal-hearing (NH) listeners. Thus, the field has focused on experimental paradigms that involve passive listening tests. However, conversation is a complex phenomenon involving not only speech production and perception but also the interaction and the adaptation of talkers to each other and their environment. The proposed research program aims to address this challenge by gaining new insights into the critical elements of interactive conversation. This research will investigate the timing of turn taking in conversation. To achieve the rapid turn taking that is typical of everyday conversation, talkers need to conduct three things in parallel: (1) understand the speech their partners are producing, (2) plan a speech response, and (3) predict when their partners will end their turn. Thus, when listening becomes more difficult (e.g., due to noise or a hearing loss), talkers may not be able to maintain the timing of turn taking that is typical of conversation in quiet. The proposed research has four objectives: (1) characterize the effect of noise and hearing loss on the prediction of the timing of turn ends in conversation; (2) characterize acoustic markers that can be used by hearing aids to estimate the difficulty listeners are experiencing during a conversation; (3) develop lab-based measures of listening effort during conversation; and (4) develop new hearing-aid processing evaluation paradigm based on measuring conversation. In the near-term, the results here will lead to new methods to better evaluate novel hearing-aid features and signal-processing algorithms. Further, understanding how turn-taking behaviour differs between normal-hearing and hearing-impaired talkers could be used to develop new communication strategies used in counselling. Through the combination of research results and the development of student training, the proposed research program will support the development of the next generation of hearing aids, significantly improving the quality of life for an aging population, increasing the productivity of workers retiring later in life, and making it easier for others to communicate with the one in 10 Canadians with a speech, language or hearing impairment. Finally, improving our fundamental understanding of interactive behaviour during conversation will also impact other areas, such as the development of more natural human-computer voice interaction systems.