An Improved Independent Component Analysis Algorithm and Its Application in Preprocessing of Bearing Sounds

Abstract

Independent Component Analysis (ICA) is known as an efficient technique to separate individual signals from various sources without knowing their prior characteristics. Firstly, the basic principle of ICA is reviewed in Sec 2, and then an improved ICA algorithm based on coordinate rotation (CR-ICA) is proposed. Secondly, two advantages of the CR-ICA algorithm are discussed; the one is that the separation can be carried out without iteration, and the other is that less computation is needed to achieve the same effect. Finally, the experiment in recognition of mixed sound and practical application in preprocessing of bearing sounds proved that the CR-ICA algorithm is better than traditional ICA algorithm in separation precision and computation speed. Moreover, the advantages of the method and the potential for further applications are discussed in the conclusion.