字典学习是信号稀疏分解研究的热点问题. 在稀疏分解字典学习中，初始字典的选择影响字典学习的效果. 为减小初始字典对学习字典的影响，在递归最小二乘(recursive least squares, RLS) 字典学习方法中引入遗忘因子的概念. 比较了最优方向法、K 奇异值分解方法和RLS 等3 种方法的字典学习效果. 分析了RLS 字典学习中不同的遗传因子对字典学习效果的影响，以及遗忘因子为不同函数时的字典学习效果. 仿真结果表明：RLS 字典学习方法减小了初始字典对学习结果的影响，故学习效果较好；而在RLS 字典学习中不同遗忘因子的选择会影响字典学习效果.
Dictionary learning is a hot topic in signal sparse decomposition. The choice of the initial dictionary affects the result of dictionary learning. In order to reduce the effects, a forgetting factor is introduced into the recursive least squares (RLS) dictionary learning. The dictionary learning effects of three dictionary
learning methods, method of optimal directions (MOD), K singular value decomposition (KSVD), and RLS are compared. Influences of different fixed forgetting factors on the final learned dictionary are analyzed,and the results of dictionary learning with different forgetting factors studied. Simulation shows that the RLS dictionary learning reduces the influence of the initial dictionary, and gives better effects. Results of the dictionary learning are influenced by the choice of the forgetting factor.
[1] 焦李成，杨淑媛，刘芳，侯彪. 压缩感知回顾与展望[J]. 电子学报, 2011, 39(7): 1651-1662. JIAO Licheng, YANG Shuyuan, LIU Fang, HOU Biao. Development and prospect of compressive sensing [J]. Acta Electronica Sinica, 2011, 39(7): 1651-1662. (in Chinese)
[2] SKRETTING K, ENGAN K. Image compression using learned dictionaries by RLS-DLA and compared with K-SVD[C]//Proc. IEEE International Conference Acoust., Speech, Signal Process (ICASSP), 2011, Pargue, Czech Republic, 2011:1517-1520.
[3] M. Elad. Sparse and redundant representations from theory to applications in signal and image processing [D]. New York: Springer, 2010.
[4] 邓承志，曹汉强. 多尺度脊波字典的构造及其在图像编码中的应用[J]. 图像图形学报，2009,14(7):1273-1278. (DENG Chengzhi, CAO Hanqiang. Construction of multi-scale ridgelet dictionary and its application for image coding[J]. Journal of Image and Graphics, 2009, 14(7):1273 -1278. (in Chinese)
[5] 孙玉宝，肖亮，韦志辉，邵文泽. 基于Gabor感知多成份字典的图像稀疏表示算法研究[J].自动化学报，2008,34(11):1379-1387. SUN Yubao, XIAO Liang, WEI Zhihui, SHAO Wenze. Sparse representations of images by a multi-component Gabor perception dictionary[J]. Acta Automatica Sinica, 2008, 34(11): 1379-1387. (in Chinese)
[6] SHEN Bin, HU Wei, ZHANG Yimin, ZHANG Yujin. Image inpainting via sparse representation [C]//Proc. IEEE International Conference Acoust., Speech, Signal Process. Taiwan, 2009: 697-700.
[7] 刘金江，王春光，孙即祥. 基于稀疏分解和神经网络的心电信号波形检测即识别[J]. 信号处理，2011,27(6):843-850. LIU Jinjiang, Wang Chunguang, SUN Jixiang. The detection and recognition of electrocardiogram’s waveform based on sparse decomposition and neural network[J]. Signal Processing,2011, 27(6): 843-850. (in Chinese)
[8] 杨清山，郭成安，金明录. 基于Gabor 多通道加权优化与稀疏表征的人脸识别方法[J]. 电 子与信息学报，2011,33(7): 1618-1624. YANG Qingshan, GUO Chengan, JIN Minglu. Face recognition based on Gabor multi- channel weighted optimization and sparse representation[J]. Journal of Electronics & Information Technology, 2011, 33(7): 1618-1624. (in Chinese)
[9] 彭富强，于德介，刘坚. 一种基于多尺度线调频基的稀疏信号分解方法[J]. 振动工程学报，2010,23(3):333-338. PENG Fuqiang，YU Dejie，LIU Jian. Sparse signal decomposition method based on multi-scale chirplet [J]. Journal of Vibration Engineering, 2010, 23(3):333- 338. (in Chinese)
[10] Karl Skretting K, Kjersti Engan K. Recursive least squares dictionary learning algorithm [J]. IEEE Transactions on Signal Processing, 2010, 58(4): 2121-2130.
[11] Michal Aharon M, Michael Elad M, Alfred Bruckstein A. K-SVD: an algorithm for designing overcomplete dictionaries for sparse representation [J]. IEEE Transactions on Signal Processing, 2006, 54 (11):4311- 4322.
[12] ENGAN K, AASE S O, HUSOY J H. Method of optimal directions for frame design [C]// Proc. IEEE International Conference Acoust., Speech, Signal Process. (ICASSP), 1999: 2443-2446.
[13] Maria G. JAFARI M G, Mark D. PLUMBLEY M D. Fast dictionary learning for sparse representations of speech signals [J]. IEEE Journal of Selected Topic in Signal Processing, 2011, 5(5):1025- 1031.
[14] Md. Mashud Hyder M M, Kaushik Mahata K. An improved smoothed approximation algorithm for sparse representation [J]. IEEE Transactions on Signal Processing, 2010, 58(4): 2194-2205.
