CIF:Small: Dynamic Dictionary Learning with Low-rank Interference

CIF:Small：具有低秩干扰的动态字典学习

• 批准号: 1422995
• 负责人: Trac Tran
• 金额: $ 26.6万
• 依托单位: Johns Hopkins University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-09-01 至 2017-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1422995&HistoricalAwards=false
• 关键词: CIF; Small; Dynamic; Dictionary; Learning

By 2020, it is projected that every human being will contribute 1,000GBof the sensor data individually. These sensors include but not limited to GPS, accelerometer, gyroscope, microphone, camera, all kinds ofwearable biomedical sensors, etc. Beyond the domain of monitoringhuman activity, multi-sensor data processing has been an activeresearch topic within the context of numerous practical applications, such as medical image analysis, remote sensing, and militarytarget/threat detection. One powerful tool to tackle these criticalBig Data problems is sparsity-driven signal-processing techniques. Asparse representation not only provides better signal compression forbandwidth/storage efficiency, but also leads to faster processingalgorithms as well as more effective signal separation for detection,classification and recognition purposes since it focuses on the mostintrinsic property of the data. Sparse signal representation allows one tocapture the hidden simplified structure present in the data jungle, andthus minimizes the harmful effects of noise in practical settings.This research investigates the problem of dynamic dictionary learning toobtain the most effective and adaptive sparse data representation evenin the presence of low-rank interference. The investigator seeks todevelop a general framework that takes advantage of diversity, yetcomplementary, features in vast correlated data sources collected frommultiple, possibly heterogeneous, sensors co-located in the same spatio-temporal physical space, recording the same physical event. Such scenarios ensure that interference noise patterns are very similar,hence justifying the low-rank property of the interference, while allowspecific structural sparsity pattern of the signal of interest to be capturedin the dictionary and/or in the sparse codes.

到2020年，预计每个人将单独贡献1,000 GB的传感器数据。这些传感器包括但不限于GPS、加速度计、陀螺仪、麦克风、摄像头、各种可穿戴生物医学传感器等。除了监测人体活动领域之外，多传感器数据处理一直是众多实际应用背景下的一个活跃研究主题，例如医学图像分析、遥感和军事目标/威胁检测。解决这些关键大数据问题的一个强大工具是稀疏驱动的信号处理技术。稀疏表示不仅提供了更好的信号压缩带宽/存储效率，但也导致更快的处理算法，以及更有效的信号分离检测，分类和识别的目的，因为它专注于最内在的属性的数据。稀疏信号表示允许人们捕捉数据丛林中隐藏的简化结构，从而最大限度地减少实际环境中噪声的有害影响，本研究探讨了动态字典学习问题，即使在存在低秩干扰的情况下，也能获得最有效和自适应的稀疏数据表示。研究人员试图开发一个通用框架，利用多样性，但互补性，功能在巨大的相关数据源收集从多个，可能异构，传感器共同位于同一时空物理空间，记录相同的物理事件。这样的场景确保干扰噪声模式非常相似，因此证明了干扰的低秩属性，同时允许在字典和/或稀疏码中捕获感兴趣的信号的特定结构稀疏模式。