Meetings

Link to page

12 June -Meeting

EEG Signal Preprocessing:

1. Filtering - Bandpass Filter

Description: Bandpass filtering removes frequencies outside a specified range. In EEG data, this can remove low-frequency drift and high-frequency noise, isolating the frequencies of interest.

Benefits:

Noise Reduction: Eliminates irrelevant frequencies that could interfere with the analysis.

Focus on Relevant Signals: Enhances the signal quality by retaining only the frequencies of interest.

Improves Model Performance: Cleaner signals lead to better feature extraction and model accuracy.

2. Re-referencing

Description: Re-referencing adjusts the EEG data to a common reference point, such as the average of all electrodes or a specific reference electrode.

Benefits:

Consistency: Creates a common baseline for all channels, reducing variability.

Noise Reduction: Can help in reducing noise common to all electrodes.

Improves Signal Quality: Enhances the comparability of signals across different channels.

3. Artifact Removal (EOG, EMG)

Description: Artifacts from eye movements (EOG) or muscle activity (EMG) can contaminate EEG signals. Independent Component Analysis (ICA) is used to identify and remove these artifacts.

Benefits:

Artifact Reduction: Removes non-brain activity signals, improving the quality of EEG data.

Cleaner Data for Analysis: Allows for more accurate analysis of brain activity.

Enhances Deep Learning Performance: Reduces the risk of models learning from irrelevant noise.

4. Normalization

Description: Normalization scales the EEG data to a standard range or distribution, often to zero mean and unit variance.

Benefits:

Standardization: Ensures that all data points contribute equally to the analysis.

Facilitates Model Training: Helps in faster convergence and better performance of deep learning models.

Comparability: Makes it easier to compare and combine data from different sources or sessions.

5. Windowing (Segmentation)

Description: Windowing is a common preprocessing technique used in time-series data, including EEG data. It involves dividing the continuous EEG signals into smaller segments or windows. Each window represents a portion of the signal over a specified period of time.

Benefits:

Feature Extraction: Windowing allows extracting features from each segment of the signal, which can be used to train machine learning models.

Time-Frequency Analysis: This technique helps in analyzing the temporal characteristics of the EEG signals and enables the application of time-frequency analysis methods like Short-Time Fourier Transform (STFT).

Handling Non-Stationarity: EEG signals are often non-stationary, and windowing helps in analyzing them as if they were stationary within each window.

Training preprocessed data:

Combination 1: Standard Preprocessing Pipeline with Non-overlapped Windows

Filtering - Bandpass Filter: 1 Hz to 40 Hz

Re-referencing: Average reference

Artifact Removal: ICA for EOG and EMG artifacts

Normalization: Z-score normalization

Windowing: 1-second windows with non-overlapping windows

Z-score Normalization: This method involves subtracting the mean and dividing by the standard deviation of the signal for each channel. This transforms the EEG signals so that they have a mean of zero and a standard deviation of one, facilitating comparisons across different conditions or subjects.

Model 1: EEGNet

Model 1: EEGNet

Subject

No preprocessing

a+ b

a + c

a + d

a + e

all

Accuracy: 62.91% Precision: 0.62 Recall: 0.63 F1-score: 0.61

Accuracy: 51.00% Precision: 0.58 Recall: 0.51 F1-score: 0.44

Accuracy: 50.06% Precision: 0.60 Recall: 0.50 F1-score: 0.40

Accuracy: 47.99% Precision: 0.41 Recall: 0.48 F1-score: 0.39

Accuracy: 49.94% Precision: 0.52 Recall: 0.50 F1-score: 0.41

Accuracy: 38.46% Precision: 0.15 Recall: 0.38 F1-score: 0.21

S2S1

Accuracy: 56.40% Precision: 0.55 Recall: 0.56 F1-score: 0.52

Accuracy: 50.96% Precision: 0.50 Recall: 0.51 F1-score: 0.46

Accuracy: 46.23% Precision: 0.61 Recall: 0.46 F1-score: 0.37

Accuracy: 44.38% Precision: 0.43 Recall: 0.44 F1-score: 0.35

Accuracy: 46.11% Precision: 0.49 Recall: 0.46 F1-score: 0.38

Accuracy: 46.15% Precision: 0.32 Recall: 0.46 F1-score: 0.37

S2S2

Accuracy: 31.25% Precision: 0.31 Recall: 0.31 F1-score: 0.31

Accuracy: 42.28% Precision: 0.42 Recall: 0.42 F1-score: 0.41

Accuracy: 41.05% Precision: 0.40 Recall: 0.41 F1-score: 0.35

Accuracy: 38.24% Precision: 0.33 Recall: 0.38 F1-score: 0.31

Accuracy: 41.05% Precision: 0.40 Recall: 0.41 F1-score: 0.35

Accuracy: 45.41% Precision: 0.47 Recall: 0.45 F1-score: 0.42

Accuracy: 40.09% Precision: 0.46 Recall: 0.40 F1-score: 0.29

Accuracy: 38.79% Precision: 0.39 Recall: 0.39 F1-score: 0.28

Accuracy: 39.18% Precision: 0.39 Recall: 0.39 F1-score: 0.29

Accuracy: 39.69% Precision: 0.46 Recall: 0.40 F1-score: 0.28

Accuracy: 57.14% Precision: 0.37 Recall: 0.57 F1-score: 0.45

Accuracy: 61.80% Precision: 0.56 Recall: 0.62 F1-score: 0.53

Accuracy: 52.47% Precision: 0.44 Recall: 0.52 F1-score: 0.39

Accuracy: 52.36% Precision: 0.40 Recall: 0.52 F1-score: 0.38

Accuracy: 52.36% Precision: 0.41 Recall: 0.52 F1-score: 0.37

Accuracy: 51.81% Precision: 0.37 Recall: 0.52 F1-score: 0.37

Accuracy: 57.14% Precision: 0.48 Recall: 0.57 F1-score: 0.523

Accuracy: 39.07% Precision: 0.28 Recall: 0.39 F1-score: 0.32

Accuracy: 44.73% Precision: 0.45 Recall: 0.45 F1-score: 0.34

All

Accuracy: 40.43% Precision: 0.16 Recall: 0.40 F1-score: 0.23

Accuracy: 42.55% Precision: 0.42 Recall: 0.43 F1-score: 0.30

There are no rows in this table

⁠

UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

Print detailed classification report

precision recall f1-score support

0 0.00 0.00 0.00 939 1 0.00 0.00 0.00 279 2 0.00 0.00 0.00 933 3 0.40 1.00 0.58 2733 4 0.00 0.00 0.00 1875

accuracy 0.40 6759 macro avg 0.08 0.20 0.12 6759 weighted avg 0.16 0.40 0.23 6759

Generated Videos

NCC

Preparation of Experiment:

⁠

https://doodle.com/meeting/organize/id/aM5n0AAa⁠

⁠

Publishers:

1) IEEE International Conference on Robotics and Automation (ICRA): - Q1

Submission of final paper: March 6, 2024

⁠

https://www.ieee-ras.org/conferences-workshops/fully-sponsored/icra⁠

⁠

2) International Conference on Intelligent Robots and Systems (IROS): -Q1

Submission of paper: March 15, 2024

⁠

https://iros2024-abudhabi.org/⁠

3) International Conference on Robotics and Automation Sciences (ICRAS): -Q2

Submission of paper: May 1st, 2024

⁠

https://www.icras.org/⁠

⁠

4) International Conference on Real-time Computing and Robotics (RCAR): -Q2

Submission of paper: March 15, 2024

⁠

http://www.ieee-rcar.org/⁠

⁠

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.