Information Representation and Computation in the Time Domain

时域信息表示与计算

• 批准号: 0635252
• 负责人: Aurel Lazar
• 金额: $ 30.18万
• 依托单位: Columbia University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-09-15 至 2010-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0635252&HistoricalAwards=false
• 关键词: Information; Representation; Computation; Time; Domain

As the critical dimensions in semiconductor technologies continue to scale into the nanoscale regime, device speeds are increasing significantly, the supply voltage is decreasing very significantly while the threshold voltage is not being reduced as aggressively. Technology scaling and the associated reduction of the supply voltageleads to a reduction of power consumption for digital ICs while still allowing the clock rate to increase. For analog front-end circuits power supply scaling is becoming a fundamental limitation. These circuits provide the indispensable interface between the analog physical world and the digital signal processing architecture. The reduction of the signal amplitude, due to drastically lower power supplies in nanoscale semiconductors, results in a very significant reduction in the accuracy of the analog circuits. A novel approach is needed for end-to-end signal processing below 1~V. Rather than being limited by the disappearing voltage headroom, the emergence of phenomenal devicespeeds has to be used. A paradigm shift is in order - instead of encoding information in the amplitude domain, time domain encoding should be employed.The research investigates a new theoretical foundation for the representation and computation of information in the time domain. Three main research themes are studied:(i) Information Representation and Signal Recovery in the Time Domain,(ii) Computation in the Time Domain, and(iii) Real-Time Algorithms and Robustness.The research extends the representation of one dimensional time dependentsignals to two and three dimensional space-time signals. By adding one or two spatialdimensions, models of speech and video processing inspired by the inner ear and the retina, two of the key natural sensory systems can be explored. The intellectual merit of this project is in devising an unique approach for computation in the time domain that is inspired by models arising in natural sensory systems and is established on soundtheoretical foundations of cutting edge mathematical modeling. The broader impacts of the proposed activity include (i) an exploration of the vast potential in the crossroads of neuroscience and nanotechnology, and (ii) an interdisciplinary undergraduate capstone lab project and a graduate level course.LEVEL OF EFFORT===============At the recommended level of support, the PI will make every effort to meet the originalscope and level of effort of the project.

随着半导体技术中的关键尺寸继续按比例缩小到纳米级范围，器件速度显著增加，电源电压非常显著地降低，而阈值电压没有被积极地降低。 技术扩展和相关的电源电压降低可降低数字IC的功耗，同时仍能提高时钟速率。 对于模拟前端电路，电源缩放正在成为一个基本的限制。这些电路提供了模拟物理世界和数字信号处理架构之间不可或缺的接口。 由于纳米级半导体中的电源大幅降低，信号幅度的降低导致模拟电路的精度显著降低。 低于1 V的端到端信号处理需要一种新的方法，而不是被消失的电压裕量所限制，必须使用惊人的器件速度的出现。一个范式的转变是为了-而不是在幅度域中编码的信息，时域编码应采用。该研究探讨了一个新的理论基础，在时域中的信息的表示和计算。 本文主要研究了三个方面的问题：（i）时域信息表示与信号恢复，（ii）时域计算，（iii）实时算法与鲁棒性，将一维时变信号的表示扩展到二维和三维空时信号。通过添加一个或两个空间维度，受内耳和视网膜启发的语音和视频处理模型，可以探索两个关键的自然感觉系统。 该项目的智力价值在于设计了一种独特的时域计算方法，该方法受到自然感觉系统中出现的模型的启发，并建立在尖端数学建模的理论基础上。 拟议活动的更广泛的影响包括：（i）在神经科学和纳米技术的十字路口的巨大潜力的探索，和（ii）跨学科的本科顶点实验室项目和研究生水平的课程。努力水平=