Energy based wavelet selection for de-noising pd signals using modified wavelet packet transform

Summary

This paper proposes a modified Wavelet Packet Transform (WPT) based denoising technique for extracting PD signals which are completely mixed with noise. WPT based de-noising technique involves four inherent steps a) decomposition using a mother wavelet, b) best basis selection for reconstruction, c) thresholding the coefficients related to the selected basis and d) reconstruction of signal from the thresholded coefficients of the selected basis. In the modified WPT based de-noising technique an energy based wavelet selection is used for selecting the mother wavelet, which is used to decompose the signal to J levels and the J is calculated using the equation J=fix(log2N). The best basis selection and thresholding procedures were eliminated by storing the decomposed coefficients in an array and introducing a frequency criterion for selecting the coefficients necessary for reconstructing the PD signal. The frequency criterion is set as per IEC 60270 with centre frequency of 250 kHz and a bandwidth of 300 kHz thereby extracting signals between 100 kHz to 400 kHz. For improving the time domain resolution of the signal the bandwidth of the signal is increased to 500 kHz, and algorithm was designed to extract signals from 0 to 500 kHz. Reconstruction was carried out by combining the coefficients selected using the above frequency criterion.

Additional informations

Authors

N.R. BURJUPATI

Keywords

de-noising PD signals, wavelet packet transform, noise elimination