Automatic correction of short-duration artefacts in single-channel EEG recording: a group-sparse signal denoising algorithm

A new algorithm for correcting short-duration artefacts in single-channel EEG recording is proposed here. A short-duration artefact has the characteristics of a group-sparse (GS) signal, i.e. a sparse signal in which its non-zero entries tend to concentrate in groups. Therefore, the problem of extracting a short-duration artefact from a recorded EEG data is formulated in the proposed algorithm as a denoising optimisation problem in which a penalty term is utilised to encourage the group sparsity of the estimated artefact. A quadratic penalty term is utilised in the proposed algorithm, and, hence, a closed-form solution is obtained in each iteration of the algorithm. In addition, a technique for automatic selection of the regularisation parameter, which controls the trade-off between the sparsity of the estimated artefact and its closeness to the measured EEG data, is also proposed. Simulation results on real and simulated EEG data show that the proposed algorithm can be successfully utilised to correct different kinds of short-duration artefacts in both single-channel and multichannel recordings. In addition, the proposed algorithm is also shown to produce significantly improved results compared to existing techniques used for performing similar tasks.