Fig. 2

Illustration of the wavelet decomposition coefficients’ rejection probability and its comparison with threshold probability. For coefficient \( w \), its corresponding rejection probability is \( p_{w} \), which is the integration from \( - \left| w \right| \) to \( \left| w \right| \) with assumed Gaussian probability density function. \( T \) is the threshold of \( w \) and \( p_{T} \) is corresponding threshold of \( p_{w} \). If \( w \) is artifact-free, the coefficients are spread around zero. Then the \( p_{w} \) will be less than \( p_{T} \). If \( w \) is contaminated with movement artifacts, the coefficients will be discretely distributed and \( p_{w} \) will be larger than \( p_{T} \)