Brain-Computer Interfaces (BCI) are systems that may function as communication channels between people and external devices through brain information. BCIs using Electroencephalography (EEG) combined with robotic systems, such as minibikes, have enabled the rehabilitation of stroke patients by decoding their actions and executing a motor task. However, the Signal-To-Noise Ratio (SNR) of EEG is low, and there is intersubject variability for pedaling detection through brain information, which reduces the Accuracy of the rehabilitation devices. Additionally, in real-Time BCIs, it is necessary to maintain a good ratio of detection and execution times. In this work, it is proposed a methodology based on an Extreme Learning Machine (ELM) to identify when the subject is executing pedaling tasks through EEG, which allows efficient detection with a maximum Accuracy of 0.85 and a False Positive Rate of 0.07. Additionally, four different frequency bands in the filtering stage were evaluated, and the results allowed concluding that the most discriminant information was available between two frequency bands: 3-7 Hz and 7-13 Hz, with an area under the ROC curve average of 0.71. The results indicate that the proposed method is suitable for the detection of pedaling tasks using EEG, which could be used for the control of a real-Time BCI for lower-limb rehabilitation.