Matching Pursuits with random sequential subdictionaries

被引:5
作者
Moussallam, Manuel [1 ]
Daudet, Laurent [2 ,3 ]
Richard, Gael [1 ]
机构
[1] Telecom ParisTech, Inst Telecom, CNRS LTCI, F-75014 Paris, France
[2] Paris Diderot Univ, ESPCI ParisTech, Inst Langevin, F-75238 Paris 05, France
[3] Inst Univ France, Paris, France
关键词
Matching Pursuits; Random dictionaries; Sparse approximation; Audio signal compression; SIGNAL MODELS; SPARSE; REPRESENTATIONS; APPROXIMATION; CONVERGENCE; ALGORITHM;
D O I
10.1016/j.sigpro.2012.03.019
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Matching Pursuits are a class of greedy algorithms commonly used in signal processing, for solving the sparse approximation problem. They rely on an atom selection step that requires the calculation of numerous projections, which can be computationally costly for large dictionaries and burdens their competitiveness in coding applications. We propose using a non-adaptive random sequence of subdictionaries in the decomposition process, thus parsing a large dictionary in a probabilistic fashion with no additional projection cost nor parameter estimation. A theoretical modeling based on order statistics is provided, along with experimental evidence showing that the novel algorithm can be efficiently used on sparse approximation problems. An application to audio signal compression with multiscale time-frequency dictionaries is presented, along with a discussion of the complexity and practical implementations. (C) 2012 Elsevier B.V. All rights reserved.
引用
收藏
页码:2532 / 2544
页数:13
相关论文
共 45 条
[31]   Matching pursuit filters applied to face identification [J].
Phillips, PJ .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 1998, 7 (08) :1150-1164
[32]   ANALYSIS SYNTHESIS FILTER BANK DESIGN BASED ON TIME DOMAIN ALIASING CANCELLATION [J].
PRINCEN, JP ;
BRADLEY, AB .
IEEE TRANSACTIONS ON ACOUSTICS SPEECH AND SIGNAL PROCESSING, 1986, 34 (05) :1153-1161
[33]  
Puy G., 2011, UNIVERSAL EFFICIENT
[34]   Union of MDCT Bases for Audio Coding [J].
Ravelli, Emmanuel ;
Richard, Gal ;
Daudet, Laurent .
IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2008, 16 (08) :1361-1372
[35]   Audio Signal Representations for Indexing in the Transform Domain [J].
Ravelli, Emmanuel ;
Richard, Gael ;
Daudet, Laurent .
IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2010, 18 (03) :434-446
[36]   Three experiential practices and their relation to psychoanalysis [J].
Schneider, Peter .
PSYCHOANALYTIC PSYCHOLOGY, 2008, 25 (02) :326-341
[37]   Efficient auditory coding [J].
Smith, EC ;
Lewicki, MS .
NATURE, 2006, 439 (7079) :978-982
[38]   Dark energy in sparse atomic estimations [J].
Sturm, Bob L. ;
Shynk, John J. ;
Daudet, Laurent ;
Roads, Curtis .
IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2008, 16 (03) :671-676
[39]  
Sturm BL, 2010, CONF REC ASILOMAR C, P581, DOI 10.1109/ACSSC.2010.5757627
[40]   Sparse Approximation and the Pursuit of Meaningful Signal Models With Interference Adaptation [J].
Sturm, Bob L. ;
Shynk, John J. .
IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2010, 18 (03) :461-472