Classification of multiclass motor imagery EEG signal using sparsity approach.

Motor imagery (MI) based brain–computer interface systems involving multiple tasks are highly required in many real-time applications such as hands and touch-free text entry, prosthetic arms, virtual reality systems, movement of a wheel chair, cursor movement, etc. The classification of MI data is the core computing in all these systems. However, the existing classification techniques are either computationally expensive or not so accurate or both. To address this limitation, in this work, a sparse representation based classification technique has been proposed to classify multi-tasks MI electroencephalogram data. The proposed method computes only wavelet energy directly from the segmented MI data and constructs a dictionary. The sparse representation from the dictionary is then used to classify given a test data. The proposed approach is faster as it works with only a single feature and without the need for any pre-processing. Further, with a reduced length of an imaging period, the proposed method provides accurate classification in a lesser computation time. The performance of the proposed approach has been evaluated and also compared with other classifiers reported in the literature. The results substantiate that the proposed sparsity approach performs significantly better than the existing classifiers.