Lọc thích nghi cho tín hiệu điện não dựa trên ước lượng bình phương tối thiểu hồi qui

Abstract

The measured brain signals are affected by various interference sources and should be removed to obtain a clean brain signal. Most research works of removing noises from brain signals have performed based on offline filtering techniques such as low-pass filtering, wavelet filtering, ICA (independent component analysis), etc., which did not meet the real time application requirements in peripheral control based on brain signals. Thus, in this research, we use an adaptive filtering algorithm based on recursive least squares estimation (RLSE) to eliminate interference for EEG signal (electroencephalography). First, a simulation of brain signal and known noise mix together. Then, use an adaptive filtering algorithm based on RLSE to reduce noise and reconstruct the simulation of brain signal. After that, this algorithm will be apply to the real EEG signal from the EPOC (Emotiv). Finally, the results will to compare with low-pass filter and wavelet filter. Experimental results showed that the adaptive filtering algorithm based on RLSE is better than low-pass filter and wavelet filter. In particular, the adaptive filtering algorithm based on RLSE can apply adaptive brain signal in real time while the wavelet filter can’t do so.