Terabit per second Data Processing to Extract Features of Interest in Enormous Amount of Data

每秒太比特的数据处理可提取海量数据中感兴趣的特征

• 批准号: 1202575
• 负责人: Alan Willner
• 金额: $ 38万
• 依托单位: University of Southern California
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2012
• 资助国家: 美国
• 起止时间: 2012-07-01 至 2017-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1202575&HistoricalAwards=false
• 关键词: Terabit; per; second; Data; Processing

Intellectual Merit: Fields as diverse as physics, environmental science, and sociology have all been dramatically transformed by the ability to utilize large amounts of data, whether it comes from computers, sensors, storage or imaging. However, the relatively "slow" speed of electronic signal processors compared to the rate at which new data is acquired is creating a bottleneck in the timely processing of data. Signal processing may involve: (a) image refining requiring signal processing on all parts of the data, and optical processing can operate with high data rate, or (b) pattern recognition requiring "scanning" (i.e., correlating) of the large data to locate a desired feature, and optical processing can potentially take Terabytes of data, rapidly identify key requested features, and then allow electronics to process Gigabytes with more detail. This proposal is to research tunable, reconfigurable complex-coefficient tapped-delay-line optical processors to enhance the processing speed for pattern recognition and data processing. The approach utilizes multiple lasers as optical pumps, a dispersive element with a wavelength-dependent propagation speed, nonlinear wave mixers, and a phase/amplitude programmable filter. All parameters of the proposed scheme (i.e., tap delays and coefficients) can be tuned by varying the power and wavelength of the pump lasers. This allows for bit-rate tunable, data-modulation-format transparent phase/amplitude processing of an optical data signal. This scheme can be used as a building block to realize more advanced functionalities, such as 2-dimentional discrete Fourier transforms.Broader Impact: This technology has the potential to directly advance many areas that use sophisticated data and image processing (e.g., environmental science, energy and sociology) by providing the capability to sift through very large data sets at faster speeds due to the high-speed ability to encode and manipulate multiple aspects of an optical wave. It can potentially enable new schemes for signal processing and reduce the time of searches within images and video data sets. This proposal hopes to utilize the large bandwidth of optics to perform digital data processing in a fully tunable and reconfigurable fashion, and to hopefully change the way Terabytes of data is processed by significantly improving the processing speed. Given the inter-connection between optics and electronics in this approach, multidisciplinary aspects can be pursued and potentially inspire novel hybrid optical/electronic architectures. Moreover: (a) high-school students will be exposed to the research, and (b) undergraduate and graduate students from diverse backgrounds will receive research training in a multidisciplinary experimental lab environment.

智力优势：物理、环境科学和社会学等不同领域都因利用大量数据的能力而发生了巨大变化，无论这些数据来自计算机、传感器、存储还是成像。然而，与获取新数据的速度相比，电子信号处理器的速度相对“慢”，这在及时处理数据方面造成了瓶颈。信号处理可能涉及：(a)图像精炼需要对数据的所有部分进行信号处理，光学处理可以以高数据速率运行，或者(b)模式识别需要对大数据进行“扫描”（即关联）以定位所需的特征，光学处理可能需要tb级的数据，快速识别关键请求的特征，然后允许电子设备处理更详细的gb级数据。本课题旨在研究可调、可重构的复系数分路延迟线光处理器，以提高模式识别和数据处理的处理速度。该方法利用多个激光器作为光泵，具有波长相关传播速度的色散元件，非线性波混频器和相位/幅度可编程滤波器。所提出方案的所有参数（即分接延迟和系数）都可以通过改变泵浦激光器的功率和波长来调谐。这允许对光数据信号进行比特率可调、数据调制格式的透明相位/幅度处理。该方案可作为实现二维离散傅里叶变换等更高级功能的基础。更广泛的影响：这项技术有可能直接推进许多使用复杂数据和图像处理的领域（例如，环境科学、能源和社会学），因为它具有对光波的多个方面进行高速编码和操作的能力，因此能够以更快的速度筛选非常大的数据集。它可以潜在地实现新的信号处理方案，并减少图像和视频数据集的搜索时间。该提案希望利用光学的大带宽以完全可调和可重构的方式执行数字数据处理，并希望通过显着提高处理速度来改变处理tb级数据的方式。鉴于这种方法中光学和电子学之间的相互联系，可以追求多学科方面，并可能激发新的混合光学/电子架构。此外：(a)高中生将接触到研究，(b)来自不同背景的本科生和研究生将在多学科实验实验室环境中接受研究训练。