A Sparse Feature Extraction Method Based on Improved Quantum Evolutionary Algorithm

Feature extraction is the key to pattern recognition. Sparse decomposition can be used to express a signal as a combination of atoms with certain structural features, being an effective way to extract the internal feature information of the signal. A sparse feature extraction method based on improved quantum evolutionary algorithm (IQEA) was proposed based on the parallelism and global search ability of IQEA to achieve fast and accurate sparse decomposition of signals on an over-complete atom dictionary. Firstly, the atoms in the dictionary were coded with probabilistic amplitudes of quantum bits, and updated by evolution-mutation cross operation of quantum bits. And then, taking the inner products of signal residual and atoms as the objective function of quantum evolution, the most characteristic atoms with signal structure were selected, and the feature extraction of signals was realized based on sparse reconstruction. Finally, the feature extraction of simulation signal and fault bearing vibration signal were carried out. The results show the effectiveness and superiority of the proposed method. © 2020, Editorial Department of Transaction of Beijing Institute of Technology. All right reserved.