Affiliations: [a] GMR Institute of Technology, Rajam, Srikakulam (Dist.), India | [b] Silicon Institute of Technology, Bhubaneswar, India | [c] S'O'A University, Bhubaneswar, India | [d] Indian Institute of Technology, Delhi, India
Correspondence:
[*]
Corresponding author: B. Biswal, GMR Institute of Technology, Rajam, Srikakulam (Dist.), 532127, A.P., India. E-mail: birendra_biswal1@yahoo.co.in
Abstract: The paper proposes a new approach for Time frequency analysis using modified time-time transform (TT-transform) for recognizing non-stationary power signal disturbance patterns. The TT-transform is derived from the well known S-transform (ST) and uses a new window function with its width inversely proportional to the frequency raised to a power 'c', varying between 0 and 1. The power disturbance signals after being processed by the TT-transform yields features, which are used for automatic recognition of disturbances; with the help of kernel based support vector machine (SVM) algorithm. Further to improve the classification performance of the TT-SVM based pattern recognizer, a differential evolution optimization algorithm (DEOA) is used. Several test cases are provided to prove the significant improvement in recognition, accuracy and drastic reduction of support vectors.
Keywords: Power signal classification, S-transform, Modified TT-transform, support vector machine (SVM), radial basis function (RBF), Mexican hat kernel function, differential evolution optimization algorithm (DEOA)
