Speech Perception Training: Advanced Scoring and Feedback Methods

言语感知训练：高级评分和反馈方法

• 批准号: 9908339
• 负责人: Silvio P Eberhardt
• 金额: $ 54.95万
• 依托单位: SEEHEAR, LLC
• 依托单位国家: 美国
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-06-01 至 2021-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9908339
• 关键词: Acoustics; Address; Adult; Age; Association Learning; Auditory; Client; Comprehension; Computer software; Computers; Control Groups; Data; Data Analyses; Data Security; Databases; E-learning; Effectiveness; Elderly; Feedback; Funding; Individual; Instruction; Label; Laboratories; Learning; Lipreading; Memory; Methods; Modality; Noise; Online Systems; Outcome; Paired-Associate Learning; Participant; Perceptual learning; Persons; Phase; Play; Publications; Randomized Controlled Clinical Trials; Randomized Controlled Trials; Recovery; Reporting; Research; Research Training; Security; Sequence Alignment; Side; Signal Transduction; Small Business Innovation Research Grant; Speech; Speech Perception; Stimulus; Sum; System; Systems Development; Testing; Text; TimeLine; Training; Verbal Learning; Visual; Vocabulary; Waiting Lists; audiovisual speech; commercialization; comparison group; design; experience; functional gain; hearing impairment; hearing threshold; improved; normal hearing; novel; operation; preservation; recruit; repository; response; social; speech in noise; speech recognition; tool; visual learning; web page; web services; word learning

Older adults commonly experience hearing difficulties in noisy settings, whether or not their hearing thresholds have increased. Being able to see the person who is talking can significantly ameliorate their hearing difficulties, but the magnitude of the benefit depends greatly on each person’s lipreading ability. Unfortunately, the vast majority of older adults who have had normal hearing during most of their lives are poor lipreaders. With Phase- I SBIR funding, SeeHear LLC realized a web-based speech recognition training system to improve lipreading. The training system implements unique high-contingency feedback following open set recognition responses during training with isolated sentences. High-contingency feedback informs the trainee about the relationship between their perceptual errors and the stimulus at the level of words or phonemes, a key to perceptual learning. Typical software for training with connected speech stimuli provides low-contingency feedback: The feedback does not establish a relationship between the stimulus and the trainee’s perceptual response errors. High- contingency feedback increases lipreading accuracy and speech recognition with audiovisual stimuli in noise in younger normal-hearing (NH) adults. This training approach will be evaluated with older hearing-impaired participants, and it will be extended beyond lipreading training to audiovisual speech recognition training in a noisy background. A second, highly novel, but potentially effective and efficient training paradigm will also be evaluated. This second approach uses a paired associates learning task, in which trainees learn to associate spoken nonsense words with nonsense pictures. The task is designed to encourage trainees to attend closely to the spoken nonsense words while trying to learn the word-to-picture associations. Laboratory testing with younger NH adults shows that lipreading training with this paradigm can deliver significant generalization to lipreading with new stimuli and a new talker. This paradigm will be tested with older adults and will be extended to offer audiovisual training in noise. The proposed project comprises randomized controlled-trial clinical tests and additional training system development. There will be a wait-list control group for comparison to groups assigned to train with sentence stimuli and high- or low-contingency feedback for open set sentence lipreading or audiovisual recognition. Two additional groups will train in the paired associates paradigm: One group will carry out training by lipreading the stimuli, and the other group will carry out training with audiovisual stimuli in speech-shaped noise. All of the participants in the research will undergo laboratory pre- and post-training tests that are designed to determine whether learning during the training sessions generalizes to different speech perception tasks, different stimuli, and different talkers across different modalities (lipreading, auditory-only, and audiovisual speech with acoustic signals in noise). In order to carry out all of the proposed training research and data recovery, the web-based system will be expanded. Successful completion of this Phase-II project will directly support commercialization of the training system.

老年人在嘈杂的环境中通常会遇到听力困难，无论他们的听力阈值是否高 能够看到说话的人可以显着改善他们的听力困难， 但好处的大小很大程度上取决于每个人的唇读能力。 大多数一生中听力正常的老年人都具有较差的唇读能力。 在 SBIR 的资助下，SeeHear LLC 实现了一个基于网络的语音识别训练系统来改善唇读。 培训系统实现了独特的高偶然性以下反馈开放集识别响应 在使用孤立句子进行训练期间，高偶然性反馈会告知受训者这种关系。 他们的知觉错误和单词或音素层面的刺激之间的关系，这是知觉学习的关键。 用于使用连接的语音刺激进行训练的典型软件提供低偶然性反馈：反馈 没有建立刺激和受训者的感知反应错误之间的关系。 应急反馈可提高唇读的准确性和噪声环境下视听刺激下的语音识别能力 年轻的听力正常 (NH) 成年人将使用老年听力障碍者来评估这种训练方法。 参与者，它将从唇读训练扩展到视听语音识别训练 第二种非常新颖但可能有效且高效的训练范式也将是。 第二种方法使用配对联想学习任务，其中受训者学习联想。 该任务旨在鼓励学员密切关注。 尝试通过实验室测试学习单词与图片的关联，同时识别口语中的无意义单词。 年轻的新罕布什尔州成年人表明，使用这种范式进行唇读训练可以显着概括 用新的刺激和新的说话者进行唇读 这种范例将在老年人中进行测试，并将得到扩展。 提供噪音视听培训。 拟议项目包括随机对照试验临床测试和额外的培训系统 将有一个等待名单对照组，用于与分配接受刑罚培训的小组进行比较。 用于开放式句子唇读或视听识别的刺激和高或低偶然性反馈 两个。 其他小组将在配对同事模式中进行培训：一组将通过唇读来进行培训 刺激，另一组将在语音形噪声中进行视听刺激训练。 研究参与者将接受实验室培训前和培训后测试，旨在确定 训练期间的学习是否可以推广到不同的语音感知任务、不同的刺激、 以及不同方式的不同说话者（唇读、纯听觉和带有声学的视听演讲） 噪声中的信号）为了进行所有建议的训练研究和数据恢复，基于网络。 该二期项目的成功完成将直接支持商业化。 的培训体系。