I. Introduction
Brain-machine interfaces (BMI) provide a new modality for communications and devices control. A number of applications are tested on human beings using non-invasive recordings from brains such as the electroencephalogram (EEG). On the other hand, direct extracellular potential recordings have been developed to extract more precise motor commands since Chapin et al [1] controlled a robotic arm using neural signals from the primary motor cortex of a rat.