Instructor: Petros T. Boufounos (email@example.com)
Schedule: TR1-2:30, Room 35-308
Modern signal and information acquisition systems employ a large variety of methods to acquire signals, and extract and represent the desired information. The increasing availability of computation enables the combination of signal and acquisition models with powerful computational methods to significantly improve the capabilities of such systems. This course explores fundamental principles as well as recent advances in the area. Topics include: Signal acquisition as a linear measurement. Basis and frame expansions for signal representations. Redundancy using overcomplete representations. Wavelets and filterbanks. Non-linear signal models. Model complexity and mean width. Sparse and structured representations. Compressive sensing and sub-Nyquist methods. Sampling and learning as function acquisition. Learning-based signal models. Learning-based recovery algorithms. Signal and information embeddings. Non-linear acquisition, time encoding, and signal-dependent sampling. Scalar and Sigma-Delta quantization. Universal quantization. Quantized Embeddings and hashing.