Dataview
LIST
FROM #Collection
WHERE file.name = this.Entry-For📗 -> 10/29/24: Signal Processing
🎤 Vocab
Dumbass pumpkin
❗ Information
Pipeline
Steps:
- Preprocessing
- Feature Extraction
- Classification
- Recognizing patterns in the data (can be ML)
Importance of Processing
- Noise Reduction and Artifact Removal
- Enhancing Signal-to-Noise Ratio
- Signal is weak compared to the noise surrounding it (muscles, etc.)
- Feature Extraction
- Frequency and Time-Frequency Analysis
- Classification and Pattern Recognition
Preprocessing: removes noise and enhances the signal of interest from the raw EEG data
Libs
MNE (Python)
- MNE (Magnetoencephalography and Electroencephalography)
Brainflow (Python)
Critical Processing
- Filtering: High-pass, low-pass, and notch filters to remove noise.
- On signal itself, not yet decomposing with Fourier
- Artifact Removal:
- Independent Component Analysis (ICA): Creates components of data
- Principal Component Analysis (PCA): Creates lower dimensional set of data. Hopes to remove irrelevant information
- Epoching: Time based epoching & Stimulation based epoching.
- Creating windows on the data
- Baseline Correction: Removing global drift.
Frequency Analysis
Time Domain -> Frequency Domain
FFT / Short-Time Fourier Transform (STFT) / Wavelet Transform / Power Spectral Density (PSD) estimation
Feature Extraction
Feature extraction is a process in which meaningful and informative attributes or features are derived from raw data.
These extracted features serve as inputs for further analysis, classification, and interpretation.
Benefits:
- Dimensionality Reduction
- Information Compression
- Machine learning
- Statistical Significance
- Discriminating Characteristics
Different Types of Features:
- Time-Domain Features
- Mean, Variance
- Hjorth Parameters
- Three measures of a signal:
- Activity (variance of signal)
- Mobility (ratio of SD of signals first derivative to the OG signal)
- Complexity (comparison of signals first and second derivatives, reflecting signal irregularity)
- Three measures of a signal:
- Peak-to-Peak Amplitude
- Difference between highest (peak) and lowest (trough) amplitude values in a time window
- Zero-Crossing Rate
- Frequency-Domain Features
- Band Power
- Peak Frequency
- Spectral Entropy
- A measure of the randomness or complexity of the signal’s frequency spectrum.
Types of Feature Extraction
- Manual feature extraction involves selecting and calculating features from data based on human expertise and domain knowledge. These features are often interpretable and tailored to specific research questions.
- Automatic feature extraction employs algorithms to derive features from data without human intervention. These algorithms learn from the data and capture complex patterns that might not be immediately apparent to humans.
Classification
- Preparing Data
- Choosing a Classifier
- Model Training
- Model Evaluation
- Fine-Tuning
Types of Classifiers
- Linear classifiers - LDA, SVM
- Neural network classifiers - CNN
- Decision Trees - Random Forest
- Nearest neighbor classifiers - KNN
SVMs - Are popular in BCI applications, especially for binary and multiclass classification tasks, like differentiating between mental states or motor imagery (imagined movement). - A discriminating ‘line’
k-Nearest Neighbors - Finds the ‘nearest’ data points. However, very sensitive to dimensionality and slow when deployed on large data sets
Training
- Learning from the data, refining its mapping from inputs to outputs
✒️ -> Scratch Notes
- Log as you go through entry
🧪-> Example
- List examples of where entry contents can fit in a larger context
🔗 -> Links
Resources
- Put useful links here
Connections
- Link all related words