Signal and system modelling: algorithms, implementations and applications

信号和系统建模：算法、实现和应用

• 批准号: 3982-2006
• 负责人: Aboulnasr, Tyseer
• 金额: $ 3.28万
• 依托单位: University of Ottawa
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2007
• 资助国家: 加拿大
• 起止时间: 2007-01-01 至 2008-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=366566
• 关键词: Signal; system; modelling; algorithms; implementations

While we may not recognize it, as people our senses never receive input in a clean, organized form.  Typically, audible signals are a mixture of conversation of several people in the same room, desirable speech mixed with background traffic noise, etc.  While we do hear the full mixture, we routinely manage to separate these signals and focus on the one that we deem important.  We are also able to shift attention quickly when a new signal arrives depending on the importance we attach to such a signal (e.g. fire alarm).  Man-made systems attempt to simulate this powerful performance of human beings in various applications.  For example, measuring the heart beat of a baby in its mother's womb will necessitate separating the fetus's heart beat from the mother's own heart beat since any sensing device will only be able to measure a mixture of both.  Similarly, microphones in a multiple speaker room will invariably pick up more than the desired speaker.  Hearing aids for a hearing impaired person in a party environment or even a kindergarten class will pick up the speech of all individuals around and some mechanism will have to determine which signals are undesirable, remove them and focus on the desired one. This research will study the signal processing aspects associated with the above problem with the objective of developing faster, lower complexity algorithms that can effectively model signals and systems with the goal of applying these algorithms to improve performance of hearing aids as one possible application.

虽然我们可能没有意识到，但作为人类，我们的感官从来没有以一种干净、有组织的形式接受过输入。通常，可听信号是同一房间中几个人的谈话，混合了背景交通噪声的理想语音等。虽然我们确实听到了完整的混合信号，但我们通常会设法将这些信号分开，并专注于我们认为重要的信号。当一个新信号到达时，我们也能够根据我们对该信号的重视程度迅速转移注意力（例如火灾警报）。人造系统试图在各种应用中模拟人类的这种强大表现。例如，测量母亲子宫内婴儿的心跳需要将胎儿的心跳与母亲自己的心跳分开，因为任何传感设备都只能测量两者的混合。同样，在一个多扬声器的房间里，麦克风总是会接收到比期望的扬声器更多的声音。在聚会环境中，甚至是在幼儿园里，为听力受损的人提供的助听器会收集周围所有人的语言，并需要一些机制来确定哪些信号是不受欢迎的，将它们删除，并将注意力集中在想要的信号上。本研究将研究与上述问题相关的信号处理方面，目的是开发更快，更低复杂度的算法，可以有效地对信号和系统进行建模，目标是将这些算法作为一种可能的应用来提高助听器的性能。