Computational Models of Hearing

人类的听觉系统能够完成语音理解、声源识别和声源分离等任务，其性能仍然是人造系统所无法比拟的。这项研究计划的长期目标是开发和利用计算模型来：i)提高我们对耳朵和大脑如何处理声学和电刺激以执行一系列听力任务的理解，ii)更好地阐明哺乳动物听觉系统中使用的神经机制，以及iii)开发改进的人类语音感知预测器和用于设计新型助听器和人工耳蜗声音处理策略的工具。这项研究计划将对在世界范围内的听力研究和语音和音频处理研究社区中广泛使用的听觉模型进行重大改进。此外，这些结果将提供对哺乳动物听觉系统中许多处理方法和生物物理机制的新理解，使其能够在如此多的任务中实现无与伦比的表现水平。这反过来将启发语音和音频处理的信号处理算法的新方向，并导致为听力受损的人提供辅助设备的进一步进步。

The human auditory system is capable of accomplishing tasks such as speech understanding, sound source identification and sound source separation at a level of performance that remains unrivalled by man-made systems. The long-term goals of this research program are to develop and utilize computational models to: i) improve our understanding of how the ear and brain process acoustic and electrical stimulation to perform a range of hearing tasks, ii) better elucidate the neural mechanisms that are utilized within the mammalian auditory system, and iii) develop improved predictors of human speech perception and tools for the design of novel hearing aid and cochlear implant sound processing strategies.This research program will achieve significant improvements in auditory models that are being utilized extensively in the worldwide hearing research and speech & audio processing research communities. In addition, the results will provide new understanding of many of the processing approaches and biophysical mechanisms employed within the mammalian auditory system that enable it to achieve its unmatched level of performance on so many tasks. This in turn will inspire new directions in signal processing algorithms for speech & audio processing and lead to further advances in assistive devices for the hearing impaired.