Robust signal processing that captures local structures by probabilistic models on dense patches

点击翻译按钮获取中文摘要

The aim of this study is to propose new multi-dimensional signal processing methods based on statistical estimation using probabilistic models on dense patches. Dense patches mean a set of voxels (neighboring voxels) that are layered with one-voxel displacement to each other. The statistics in each patch is robust against random variation and can capture local structures of brain-images. These properties contribute elastic regularization that is different from place to place. Three-dimensional signal processing based on dense-patch information has possibility to be applied for brain-machine interface engineering and neuroscience. This study has been involved to brain signal processing, decoding or BMI, and image signal processing.