According to recent researches, active training that induces voluntary motor drive has a positive impact on motor encoding. The self-paced treadmill allows users to adjust belt speed based on their movements and it may encourage voluntary movement to the user. So, the self-paced treadmill is expected to be more effective than the conventional fixed speed treadmill. Our aim was to compare neural activity while participants followed targeted fast and targeted slow speeds on two treadmill conditions, self-paced treadmill and fixed speed treadmill. We used noninvasive electroencephalography(EEG) and a force measuring treadmill to record neural activity and body dynamics during a treadmill walking. The preprocessing method rejected bad channels that indicated by kurtosis and re-referenced the EEG to a common average of the remaining channels. The processing method including the independent component analysis was used to remove the electrode, muscle, ocular artifacts and extract components that represent brain source. Cortical components are clustered across the subjects based on event-related potential(ERP) and power spectra by k-mean clustering algorithm. Finally, we examined mu, beta and gamma power in the motor cortices to observe the difference in neural activity of cortical components on different treadmill walking.