Investigation of Phase Features of Movement Related Cortical Potentials for Upper-Limb Movement Intention Detection

The movement related cortical potential (MRCP) is a well-known neural signature of humans self-paced movement intention, which can be exploited by future rehabilitation robots. Most existing studies have explored the amplitude representation for the detection. In this paper we have investigated the phase representation for such a task. On the data sets in which 15 healthy subjects executed a self-initiated upper limb center-out reaching task, we have evaluated the detection models with MRCP amplitude features, MRCP phase features and a concatenation of MRCP amplitude and phase features, respectively. The experimental results have demonstrated that the detector based on the concatenation of amplitude and phase features has not only attained the largest percentage of correct classified trials among the three models (88.05% ± 8.80% of trials), but also achieved the earliest detection of the upper-limb movement intention before the actual movement onset (634.58 ± 211.12 ms before the movement onset).