An undecimated wavelet transform based enhancement, statistical feature extraction and detection-classification of PD signals [partial discharges]

We address the problem of recognition and retrieval of relatively weak industrial signals such as partial discharges (PD) buried in excessive noise. The major bottleneck being the recognition and suppression of stochastic pulsive interference (PI) which has similar frequency characteristics as the PD pulse. In this paper, we provide techniques to de-noise, detect, estimate and classify the PD signal in a statistical perspective. A multi-resolution analysis based technique is incorporated to discard the huge amount of redundant data in the acquired signal. A scale dependent MMSE based estimator is implemented in the undecimated wavelet transform (UDWT) domain to enhance the noisy signal. We characterize the PD and PI pulses using a statistical model as the first moment of multi variate Gaussian distribution and its parameters are estimated using maximum likelihood (ML) and maximum aposteriori probability (MAP) based techniques. A statistical test, known as the generalized log likelihood ratio test (GLRT), was incorporated to ensure the existence of the pulse. The decision as to whether a pulse is a noise or a desired signal has been made based on a weighted-nearest neighbor methodology.