Although MP-based features find representative time–frequency atoms, calculating the mean and standard deviations of atom parameters eliminates the distinctiveness of the selected atoms. The MP-based features are used as supplement to Mel-frequency cepstral coefficients. Mean and standard deviations of frequency and scale parameters of selected Gabor atoms are used to produce the four-dimensional (4D) feature. Gabor atoms are adopted to build the dictionary in their work. Chu et al. introduced MP to sparse coding for environmental sound recognition. Given the dictionary, methods have been developed to find the weight vector such as matching pursuit (MP), orthogonal MP, gradient pursuit, basis pursuit and the least absolute shrinkage and selection operator . The parameters of the selected atoms indicate the properties of transient components. The atoms associated with non-zero weights are called selected atoms. (1)where z is the signal, e is the residual, D is the dictionary containing N atoms and x is the N-dimensional weight vector containing L non-zero values ( L < < N). IET Generation, Transmission & Distribution.IET Electrical Systems in Transportation.IET Cyber-Physical Systems: Theory & Applications.IET Collaborative Intelligent Manufacturing.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |