Abstract
Non-invasive EEG-based brain-computer interfaces (BCI) for handwriting imagery can support the restoration of fine writing abilities in individuals with motor impairments. However, the development of high-performance decoding algorithms is constrained by scarce training datasets. To address this, we present the first open EEG dataset dedicated to handwriting imagery. The dataset comprises 32-channel EEG recordings (sampled at 1000 Hz) from 21 healthy participants across two sessions separated by at least 24 hours. A dual-paradigm design captures multidimensional neural features: a Chinese character stroke imagery task (five basic strokes, 200 trials per session) and a Pinyin single-vowel imagery task (six vowels, 240 trials per session). After rigorous quality screening, 18,480 standardized trials are provided, ensuring completeness, reliability, and adherence to the Brain Imaging Data Structure (BIDS) standard. This dataset enables the development and evaluation of algorithms for non-invasive BCI and supports research on restoring writing-based communication in individuals with motor impairments.