CRCNS: Avian Model for Neural Activity Driven Speech Prostheses

理解人类发声的物理、计算和理论基础 交流，言语，对提高对声音、言语和语言的理解至关重要 疾病和障碍，并改进其诊断、治疗和预防。满足这一要求 挑战需要了解发声运动控制的神经和感觉运动机制。 我们的项目将直接研究神经和感觉运动机制涉及的 产生复杂、自然、有声的通信信号。我们的结果将直接提高 脑-机接口技术用于通信，将加速 用于有通讯功能的个人的假肢和其他辅助/增强技术 由于受伤或疾病造成的缺陷。我们将开发一种直接翻译神经的发声假体 皮质感觉运动区和发声运动控制区的信号转化为发声通信信号 实时输出。在成功使用非人类灵长类动物作为脑机接口的基础上 对于一般的运动控制，假肢将在鸣禽身上开发，它们的声学丰富， 习得的发声与人类说话有许多共同之处。因为鸣鸟的歌声 器官在功能和解剖学上类似于人类的喉部和皮质区 控制它的区域与人类大脑的言语运动控制区非常相似，鸣禽 为所提出的研究提供一个理想的模型。超越了我们的工作对人类声音的应用 和语音，发声假体的发展将使新的语音相关研究在 可以揭示发声学习和发声的基本机制的鸣鸟模型。在 在项目的第一阶段，我们收集了大量同时记录的神经活动和 发声。在第二阶段，我们将应用机器学习和人工智能技术来 开发将神经录音映射到声音输出的算法，使我们能够估计预期的 直接从神经数据发声。在第三阶段，我们将开发计算基础设施，以 实时运行这些算法，根据神经活动预测预期发声 动物正在积极地产生这些发声。在第四阶段，我们会测试 通过用我们的假肢系统的输出代替鸟自己的发声来进行假肢。 这些技术的成功将为在人体上测试这些技术并将其翻译成多个 辅助设备。除了我们的研究目标外，该项目还将邀请毕业生、 本科生和高中生通过开发新颖的教育模块 向学生介绍脑机接口和跨学科研究 工程学和基础科学。

Understanding the physical, computational, and theoretical bases of human vocal communication, speech, is crucial to improved comprehension of voice, speech and language diseases and disorders, and improving their diagnosis, treatment and prevention. Meeting this challenge requires knowledge of the neural and sensorimotor mechanisms of vocal motor control. Our project will directly investigate the neural and sensorimotor mechanisms involved in the production of complex, natural, vocal communication signals. Our results will directly enhance brain-computer interface technology for communication and will accelerate the development of prostheses and other assistive/augmentative technologies for individuals with communications deficits due to injury or disease. We will develop a vocal prosthetic that directly translates neural signals in cortical sensorimotor and vocal-motor control regions into vocal communication signals output in real-time. Building on success using non-human primates for brain computer interfaces for general motor control, the prosthetic will be developed in songbirds, whose acoustically rich, learned vocalizations share many features with human speech. Because the songbird vocal apparatus is functionally and anatomically similar to the human larynx, and the cortical regions that control it are closely analogous to speech motor-control areas of the human brain, songbirds offer an ideal model for the proposed studies. Beyond the application of our work to human voice and speech, development of the vocal prosthetic will enable novel speech-relevant studies in the songbird model that can reveal fundamental mechanisms of vocal learning and production. In the first stage of the project, we collect a large data set of simultaneously recorded neural activity and vocalizations. In stage two, we will apply machine learning and artificial intelligence techniques to develop algorithms that map neural recordings to vocal output and enable us to estimate intended vocalizations directly from neural data. In stage three, we will develop computing infrastructure to run these algorithms in real-time, predicting intended vocalizations from neural activity as the animal is actively producing these vocalizations. In stage four, we will test the effectiveness of the prosthetic by substituting the bird’s own vocalization with the output from our prosthetic system. Success will set the stage for testing of these technologies in humans and translation to multiple assistive devices. In addition to our research goals, the project will engage graduate, undergraduate, and high school students through the development of novel educational modules that introduce students to brain machine interface and multidisciplinary studies that span engineering and the basic sciences.