Different Signal Processing Techniques used in Artifact removal from Electroencephalography (EEG) Signal.pptx

This slide is about different signal processing techniques used in artifact removal from electroencephalography (EEG) signal. It introduces three techniques: independent component analysis (ICA), wavelet transform (WT), and empirical mode decomposition (EMD). ICA is a signal processing method to separate independent sources that are linearly mixed in several sensors. WT is a mathematical technique used for analyzing signals and data in various fields, including signal processing, image processing, and data compression. EMD is designed to analyze non-stationary and nonlinear data, it is a good fit for these signals because EEG signals frequently have complex and time-varying characteristics. When artifacts taint EEG recordings, EMD can decode the signal into its constituent intrinsic mode functions (IMFs) based on the regional scales and frequencies of those IMFs.

1. Refath Artifact Removal from EEG Signal Page 1 /26 Presented by: Sojib Ahmed Refath (2030936). Partho Prosad (2022371). Department of Electrical & Electronic Engineering INDEPENDENT UNIVERSITY, BANGLADESH EEE465 PROJECT PRESENTATION Different Signal Processing Techniques used in Artifact removal from Electroencephalography (EEG) Signal Supervisor: Dr. Md. Kafiul Islam

2. Refath Artifact Removal from EEG Signal Page 2 /26 Independent Component Analysis (ICA) Wavelet Transform (WT) Empirical Mode Decomposition (EMD) Outline

3. Refath Artifact Removal from EEG Signal Page 3 /26 What is ICA? ICA is a signal processing method to separate independent sources that are linearly mixed in several sensor. Example:

4. Refath Artifact Removal from EEG Signal Page 4 /26 Compare With EEG Extraction

5. Refath Artifact Removal from EEG Signal Page 5 /26 What is EEGLAB? EEGLAB (EEG Laboratory) is a widely used open-source MATLAB toolbox for processing and analysing electroencephalography (EEG) and other electrophysiological data. It was developed by researchers at the Swartz Center for Computational Neuroscience at the University of California, San Diego (UCSD).

6. Refath Artifact Removal from EEG Signal Page 6 /26 EEGLAB has built-in functions for applying ICA to EEG data. ICA is particularly useful for separating independent brain sources from mixed EEG signals and can be helpful in removing artifacts. Removing Artifacts using ICA in EEGLAB:

7. Refath Artifact Removal from EEG Signal Page 7 /26 In our project we have been given a 14 channel EEG Data which has sampling frequency of 128 Hz that means the maximum allowable frequency is 64 Hz . Plotted the raw signal. Removing Artifacts using ICA in EEGLAB (Cond..)

8. Refath Artifact Removal from EEG Signal Page 8 /26 Preprocessing Before applying for ICA we pre-processed the signal by removing dc offset ,filtering and removing power line noise .

9. Refath Artifact Removal from EEG Signal Page 9 /26 Before and After Preprocessing(Frequency Domain) Before preprocessing After preprocessing

10. Refath Artifact Removal from EEG Signal Page 10 /26 Separations by ICA Now we are ready to decomposed by ica that is built in in eeglab . In EEGLAB, the " runica " function is used to perform Independent Component Analysis (ICA) on EEG data. When we use the " runica " function with the 'infomax' algorithm, it employs the Infomax ICA algorithm, which is one of the popular and widely used ICA algorithms for separating independent components from mixed EEG signals.

11. Refath Artifact Removal from EEG Signal Page 11 /26 After ICA Separation (Finding bad channel) After analyzing the data of 14 independent components we can observe that channel 1 and 3 has the highest abrupt changes in the graph .these may be eye blink and maybe unknown artifacts.

12. Refath Artifact Removal from EEG Signal Page 12 /26 Removing the bad channel and plot the new signal: The blue line signal is before removal of bad signal and red one after removing of bad channel

13. Refath Artifact Removal from EEG Signal Page 13 /26 Frequency domain of clean signal: Fig. Frequency Spectra After Removing Bad Channel.

14. Partho Artifact Removal from EEG Signal Page 14 /26 Wavelet Transform (WT): Wavelet transform is a mathematical technique used for analyzing signals and data in various fields, including signal processing, image processing, and data compression. Fig. EEGData Channel-01

15. Partho Artifact Removal from EEG Signal Page 15 /26 DC Offset Remove Before DC Offset Remove After DC Offset Remove

16. Partho Artifact Removal from EEG Signal Page 16 /26 Various kind of wavelet Fig. Various kind of wavelet Wavelet Decomposition ( Process)

17. Partho Artifact Removal from EEG Signal Page 17 /26 Wavelet selection and Plotting Coefficient Fig. Wavelet selection Wavelet Families

18. Partho Artifact Removal from EEG Signal Page 18 /26 D ecomposition and Reconstruction %% Wavelet Decomposition [ swa,swd ] = swt (EEG_01, 3, 'sym5' ); Reconstruction

19. Partho Artifact Removal from EEG Signal Page 19 /26 Empirical Mode Decomposition (EMD): EMD is designed to analyze non-stationary and nonlinear data, it is a good fit for these signals because EEG signals frequently have complex and time-varying characteristics. When artifacts taint EEG recordings, EMD can decode the signal into its constituent intrinsic mode functions (IMFs) based on the regional scales and frequencies of those IMFs.

20. Partho Artifact Removal from EEG Signal Page 20 /26 DC Offset Remove

21. Partho Artifact Removal from EEG Signal Page 21 /26 DC Offset Remove and IMF in pwelch

22. Partho Artifact Removal from EEG Signal Page 22 /26 Applying EMD and IMF Plotting

23. Partho Artifact Removal from EEG Signal Page 23 /26 N ew IMF Plotting and Thresholding

24. Partho Artifact Removal from EEG Signal Page 24 /26 Removing Offset from RESIDUAL, Reconstruction Fig. Reconstruction

25. Partho Artifact Removal from EEG Signal Page 25 /26 L imitation of EMD EMD is applicable for single channel. If we had applied EMD to the entire data set one by one, we would have seen 14 - channel such reconstructed datasets. While EMD has its merits, it also has several disadvantages. EMD can be computationally complex, particularly for lengthy and intricate time series data. Due to the iterative nature of the algorithm and the requirement to compute the extrema and envelopes for each IMF, processing time can be quite long.

26. Partho Artifact Removal from EEG Signal Page 26 /26 Any questions, comments or suggestions?