This dataset comprises data from six experimental participants, each undergoing nine walking trials. Each participant engaged in three trials of low-speed walking, three trials of medium-speed walking, and three trials of high-speed walking. The dataset includes multi-channel electromyography (EMG) data and center of pressure/ground reaction force (COP/GRF) data. Specifically, EMG data is utilized to extract muscle coordination activation time coefficients during human walking, and a deep learning model is established based on these coefficients to predict COP/GRF parameters.