A new multimodal dataset combining functional near-infrared spectroscopy (fNIRS) and electroencephalography (EEG) has been introduced for unilateral limb motor imagery (MI) research. Brain-computer interfaces (BCIs) acquire and decode brain activity signals to translate them into control commands for external devices, and can be divided into invasive and non-invasive types. Motor imagery is a key paradigm in non-invasive BCIs, with applications including upper-limb functional reconstruction, daily assistive control, and continuous trajectory guidance. MI involves a distributed motor–somatosensory–parietal network including the primary motor cortex (M1), primary somatosensory cortex (S1), lateral premotor cortex (PMd/PMv), supplementary motor area (SMA), and the superior parietal lobule / intraparietal sulcus (SPL/IPS). Traditional MI paradigms focus on binary- or four-class bilateral limb tasks that are well separated in spatial location, relying on hemispheric lateralization differences and requiring relatively low spatial resolution. However, application scenarios such as upper-limb stroke rehabilitation, unilateral exoskeleton control, or prosthetic control require higher-dimensional control commands, and traditional paradigms provide limited classes and control degrees of freedom. Unilateral MI with multiple directions and multiple joints offers higher control freedom and better action separability without substantially increasing cognitive load. Unlike bilateral tasks, unilateral multi-class MI does not show strong cross-hemisphere contrasts; instead, class differences are reflected by subtle spatial modulations within the