Correlated Noise Reduction Algorithm for Multichannel Leak Signal

Wavelet transform is well know by its sparse representation of signals, with application to separate the noise from signals. Wavelet transform is appropriate to multichannel signal denoising, we consider the spatially correlation between noises and develop a new correlated noise reduction algorithm. This method is based on a change of basis that is obtained from the diagonalization of a robust estimator of the noise covariance matrix. Classical test signals were simulated to analysis the performance of the algorithm. Experiments show it outperforms the primitive wavelet transform on the one dimentional signal situations. Four channel signals are obtained from the oil pipeline leak detection project. We apply multivariate denoising to these signals and this provide denoised signals for the next step of leak recognize and time delay estimation.