Career: Bilinear Shape-Constrained Regression in Blind Source Separation/Equalization, and Signal Processing for Chromatographic Analysis

职业：盲源分离/均衡中的双线性形状约束回归，以及色谱分析的信号处理

• 批准号: 9733540
• 负责人: Nikolaos Sidiropoulos
• 金额: $ 20万
• 依托单位: University of Virginia Main Campus
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-07-01 至 2000-02-29
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9733540&HistoricalAwards=false
• 关键词: Career; Bilinear; Shape; Constrained; Regression

The explosive growth of wireless communications has generated competitive pressure to fulfill the promise of seamlessly integrated personal communication services. Meeting quality-of-service constraints for advanced multimedia services over wireless is an important challenge. Wireless channels are rapidly-varying, rate-limited, and subject to severe degradation due to propagation and multiuser interference. This has sparked considerable interest in blind signal separation techniques, capable of separating and estimating the users' signals without assuming knowledge of the channel. Interestingly, mathematically similar problems appear in chromatographic analysis (CA) and flow injection analysis (FIA) in chemometrics, with applications in quality control for manufacturing, a Federal Strategic Area. The research component of this program involves the study of blind signal separation problems, posed as a (system of) bilinear regression(s) subject to: (i) factorization constraints; (ii) uni- or oligo-modality, convexity, or support constraints on the columns of certain factors (these exploit direction-of-arrival (DOA) diversity afforded by antenna arrays, without requiring DOA estimation; unimodality is well-motivated in the context of CA/FIA, and so is smoothness of factor profiles); and (iii) modulation-induced (e.g., finite-alphabet) constraints. The emphasis is on the development and performance analysis of associated optimization algorithms, and applications thereof in communications and chemometrics. Cross-fertilization between the two application domains is an integral goal of this program. The educational component of this program includes: (i) the development of introductory and advanced courses on optimization theory and algorithms for signal processing/communications -oriented undergraduate and graduate students; (ii) the development of a suite of MATLAB routines for laboratory/Web instruction; (iii) the creation of undergraduate signal processing research and education opportunities; and (iv) greater emphasis on innovative instruction through proper utilization of multimedia tools to complement traditional modes of instruction.

无线通信的爆炸性增长产生了履行无缝集成个人通信服务承诺的竞争压力。满足无线高级多媒体服务的服务质量限制是一个重要的挑战。无线信道是快速变化的、速率受限的，并且由于传播和多用户干扰而受到严重退化。这引发了人们对盲信号分离技术的相当大的兴趣，该技术能够在不假设信道知识的情况下分离和估计用户的信号。有趣的是，在化学计量学的色谱分析(CA)和流动注射分析(FIA)中出现了数学上类似的问题，应用于制造业的质量控制，这是联邦战略领域。本程序的研究部分涉及盲信号分离问题的研究，假设为满足以下条件的双线性回归(S)系统：(I)因式分解约束；(Ii)对某些因素列的单模或寡模、凸性或支持约束(这些利用了天线阵列提供的到达方向(DOA)分集，而不需要DOA估计；在CA/FIA的背景下，单峰是很有动机的，并且因子分布的平滑性也是如此)；以及(Iii)调制诱导的(例如，有限字母表)约束。重点是相关优化算法的发展和性能分析，以及它们在通信和化学计量学中的应用。这两个应用领域之间的交叉受精是该计划的一个整体目标。该计划的教育部分包括：(I)为面向信号处理/通信的本科生和研究生开发关于最优化理论和算法的入门和高级课程；(Ii)开发一套用于实验室/网络教学的MatLab例程；(Iii)创造本科生信号处理研究和教育机会；以及(Iv)通过适当利用多媒体工具来补充传统教学模式，更加强调创新教学。