For handling human-machine interactions, in this last years, many efforts have been devoted to Brain-Computer Interface (BCI). Electroencephalogram (EEG) electrodes enhance the convenience and wearability of BCI. Unfortunately, the noise induced by sampling reduces the signal quality compared to that of electrodes. In this paper a methodology for EEG waves compression and noise reduction is introduced. The approach is based on a non-linear Principal Component Analysis Neural Network for compression and decompression (reconstruction) of the data. Experiments are made on a corpus containing the activation strength of the fourteen electrodes of an EEG headset for eye state prediction. The experimental results highlight that the technique permits to obtain an higher rate of classification accuracy w.r.t. the use of row data.
Non-linear PCA Neural Network for EEG Noise Reduction in Brain-Computer Interface
Ciaramella A.
;Dezio G.;
2020-01-01
Abstract
For handling human-machine interactions, in this last years, many efforts have been devoted to Brain-Computer Interface (BCI). Electroencephalogram (EEG) electrodes enhance the convenience and wearability of BCI. Unfortunately, the noise induced by sampling reduces the signal quality compared to that of electrodes. In this paper a methodology for EEG waves compression and noise reduction is introduced. The approach is based on a non-linear Principal Component Analysis Neural Network for compression and decompression (reconstruction) of the data. Experiments are made on a corpus containing the activation strength of the fourteen electrodes of an EEG headset for eye state prediction. The experimental results highlight that the technique permits to obtain an higher rate of classification accuracy w.r.t. the use of row data.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.