Interference-driven adaptation in sparse approximation

Sparse approximation attempts to find an efficient signal representation by adaptively building a signal vector space from elements of a usually redundant and over-complete dictionary of atoms. Often, however, representations produced by iterative descent methods, such as orthogonal matching pursuit (OMP), will contain atoms that are poorly chosen and might later be confused as features of the signal. Poorly selected atoms bring about the selection other atoms that serve to correct for previous choices using destructive interference. This behavior diminishes the efficiency of a representation. In this paper, we propose and study a modification of the atom selection in OMP that takes into account the aforementioned effects. We find that a pursuit adapting to the interference between atoms can create a more efficient representation than that created by OMP, and one that is more representative of the signal and its features and less a reflection of the decomposition process.

[1]  Bob L. Sturm,et al.  Dark Energy in Sparse Atomic Estimations , 2008, IEEE Transactions on Audio, Speech, and Language Processing.

[2]  Mads Græsbøll Christensen,et al.  Efficient parametric coding of transients , 2006, IEEE Transactions on Audio, Speech, and Language Processing.

[3]  Michael Elad,et al.  Image Denoising Via Sparse and Redundant Representations Over Learned Dictionaries , 2006, IEEE Transactions on Image Processing.

[4]  Stéphane Mallat,et al.  Matching pursuits with time-frequency dictionaries , 1993, IEEE Trans. Signal Process..

[5]  Emmanuel J. Candès,et al.  Robust uncertainty principles: exact signal reconstruction from highly incomplete frequency information , 2004, IEEE Transactions on Information Theory.

[6]  A. Bruckstein,et al.  K-SVD : An Algorithm for Designing of Overcomplete Dictionaries for Sparse Representation , 2005 .

[7]  Y. C. Pati,et al.  Orthogonal matching pursuit: recursive function approximation with applications to wavelet decomposition , 1993, Proceedings of 27th Asilomar Conference on Signals, Systems and Computers.

[8]  Pascal Frossard,et al.  Low-rate and flexible image coding with redundant representations , 2006, IEEE Transactions on Image Processing.

[9]  Emmanuel Vincent,et al.  Instrument-Specific Harmonic Atoms for Mid-Level Music Representation , 2008, IEEE Transactions on Audio, Speech, and Language Processing.

[10]  C.-C. Jay Kuo,et al.  Environmental sound recognition using MP-based features , 2008, 2008 IEEE International Conference on Acoustics, Speech and Signal Processing.

[11]  K. Blinowska,et al.  Micro- and macrostructure of sleep EEG , 2006, IEEE Engineering in Medicine and Biology Magazine.

[12]  A. Willsky,et al.  HIGH RESOLUTION PURSUIT FOR FEATURE EXTRACTION , 1998 .

[13]  Laurent Daudet,et al.  Sparse and structured decompositions of signals with the molecular matching pursuit , 2006, IEEE Transactions on Audio, Speech, and Language Processing.

[14]  Michael A. Saunders,et al.  Atomic Decomposition by Basis Pursuit , 1998, SIAM J. Sci. Comput..

[15]  Martin Vetterli,et al.  Atomic signal models based on recursive filter banks , 1997, Conference Record of the Thirty-First Asilomar Conference on Signals, Systems and Computers (Cat. No.97CB36136).

[16]  M. Elad,et al.  $rm K$-SVD: An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation , 2006, IEEE Transactions on Signal Processing.

[17]  Karthikeyan Umapathy,et al.  Discrimination of pathological voices using a time-frequency approach , 2005, IEEE Transactions on Biomedical Engineering.

[18]  J.J. Shynk,et al.  A Short-Term Measure of Dark Energy in Sparse Atomic Estimations , 2007, 2007 Conference Record of the Forty-First Asilomar Conference on Signals, Systems and Computers.