This study investigates the feasibility and efficacy of decoding lower limb movement speed through the examination of differences between motor imagery and relaxation states. Electroencephalography (EEG) signals are utilized as the input data source, and commonly used machine learning approaches are employed for classifying imagined lower limb movement speed. Healthy individuals without lower limb motor impairments participate in the experiment, and their EEG signals are recorded using Emotive’s 32-channel gel electrode EEG cap EPOC FLEX. Preprocessing and feature extraction techniques are applied to the collected EEG data to develop a specialized classification model. Results indicate significant differences in EEG signals between imagined lower limb movement speed and relaxation states. Ten-fold cross-validation confirms the reliability and accuracy of the classification model, achieving above-chance classification accuracies. The findings provide valuable insights for the development of brain-computer interface systems, rehabilitation therapies, and applications related to lower limb movement. This study establishes a foundation for further exploration in decoding lower limb movement speed.