A method of estimating the positions and orientations of activated muscle fibers with electrode arrays is proposed by selecting differential channels with high signal-to-noise ratios. A finger tapping experiment is carried out to prove its effectiveness. The estimation results are consistent with the anatomical structures of the muscles. The extensors of thumb, index and middle fingers can be distinguished. Different types of electrode mini-grids are tested and compared, of which the X-shaped grid is determined to be the optimal mini-grid arrangement. This method could be applied to neuro-muscle mapping in sEMG decoding algorithms.