Combination of NMF and IVA for improving the quality of EOG signals
Wenhui Sun, Bing Xue, Jingyi Xue, Shengli Zhang, Zhao Lv · 2020
Eye movements play an important role in the implementation of human computer interface (HCI) system, which could be recorded by the electrooculography (EOG) technique. However, the recorded EOG signals are always contaminated by various latent artifacts. Aiming at obtaining high-quality EOG signals, a blind source separation (BSS) method combined with non-negative matrix factorization (NMF) and independent vector analysis (IVA), is used in the separation problem of EOG signals. The effectiveness of the addition of NMF is validated on the classification tasks, and the corresponding average classification accuracy is higher than that of extended Infomax independent component analysis (ICA) and IVA, with the improvement of 4.42% and 3.26%, respectively. This result indicates that the combination of NMF and IVA can achieve better separation performance, in which the NMF could provide more information close to the source signals in the iteration step.