作者: Mahdi Ataee , Hadi Zayyani , Massoud Babaie-Zadeh , Christian Jutten
DOI: 10.1109/ICASSP.2010.5495278
关键词:
摘要: In this paper, we suggest to use a steepest descent algorithm for learning parametric dictionary in which the structure or atom functions are known advance. The of atoms allows us find direction parameters instead itself. We also thresholded version Smoothed-l0 (SL0) sparse representation step our proposed method. Our simulation results show that using similar Gabor and these Gabor-like yield better representations noisy speech signal than non methods like K-SVD, terms mean square error representations.