This function removes noise from signals using wavelet transform. The noisy signal is first decomposed using multi-level wavelet decomposition. Then some of its detail coefficients are thresholded. Finally, the approximation coefficients and altered detail coefficients are used to reconstruct the signal.
Two methods are available to determine the threshold. They are custom and sqtwolog. If the former is chosen, you can specify the threshold used in every level by setting the value for Threshold of every level (%). A larger threshold will mean that more noise is likely to be removed, but the useful signal is also more likely to be distorted. On the contrary, a small threshold is less likely to distort the useful signal. But it might not be capable enough to remove the noise. If sqtwolog is chosen, the threshold will be automatically calculated.
If the input comes from a graph, the denoised signal will be added to the input graph.
Topics covered in this section: