A PCA Based Artifact Removal Algorithm for Neural Signal Acquisition with kS/s Sampling Rate.

The contamination of stimulus artifacts during Deep Brain Stimulation (DBS) brings challenges to the signal processing, especially when the ratio of the kS/s sampling rate to the stimulation frequency is not an integer. In this work we study to deal with this problem. A transfer function is built to describe the relationship between the stimulation signal and the artifact at the acquisition site. A principal component analysis (PCA) based linear regression algorithm for eliminating the artifact is proposed. The algorithm can be used for the artifact removal with low sampling rate of the neural signal. Higher than 60% correlation coefficient of the artifact-free signal and the predetermined self-generated signal is achieved when the artifact is 60dB larger than the predetermined signal. The numerical recipe for the critical algorithm is also proposed, lowering the complexity from cubic degree to square degree.