Skip to content

Line Noise Removal

The Line Noise Removal step allows you to remove powerline interference (50/60 Hz) and its harmonics from your HD-sEMG signals. This step is crucial for improving signal quality before further analysis.

Overview

In this step, you can: 1. Choose from multiple line noise removal methods 2. Select the appropriate powerline frequency region (US/EU) 3. Process multiple files with visual progress tracking 4. Compare different filtering approaches

Requirements

Before using the Line Noise Removal step: - Files must have been processed through the Grid Association step - The workspace must be properly initialized - For MATLAB-based methods: MATLAB Engine for Python must be installed - For CleanLine: MATLAB + EEGLAB + CleanLine plugin must be installed - For Octave: Octave and oct2py must be installed

Available Methods

1. MNE-Python: Notch Filter (FIR)

Type: Finite Impulse Response filter Speed: ⚡⚡⚡ Very fast Quality: ⭐⭐⭐ Good

Classic notch filter approach that creates narrow rejection bands at specified frequencies.

Advantages: - Very fast and efficient - Stable filtering (no phase shift with zero-phase) - Simple to understand - No external dependencies (only MNE-Python)

Disadvantages: - Also removes frequencies near the target ("frequency hole") - Not adaptive - uses fixed frequencies - Can cause time-domain distortions

Best for: Fast processing when slight spectral distortions are acceptable.

Type: Adaptive spectrum fitting with sinusoidal regression Speed: ⚡⚡ Medium Quality: ⭐⭐⭐⭐⭐ Excellent

Adaptively estimates and removes sinusoidal components using sliding windows. Similar approach to CleanLine.

Advantages: - Adaptive - adjusts to time-varying interference - Minimal distortion of adjacent frequencies - Narrower removal than classical notch filters - No external dependencies (only MNE-Python) - Similar quality to CleanLine without MATLAB

Disadvantages: - Slower than simple notch filter - More computationally intensive for long signals

Best for: High-quality signal processing with minimal distortion. This is the recommended default method for most users.

3. MATLAB CleanLine (EEGLAB Plugin) 🏆 GOLD STANDARD

Type: Adaptive multi-taper regression with Thompson F-statistic Speed: ⚡ Slow Quality: ⭐⭐⭐⭐⭐ Excellent (Gold Standard)

The original CleanLine algorithm from EEGLAB. Uses multi-taper spectral analysis with statistical validation.

Advantages: - Gold standard for adaptive line noise removal - Statistical validation using Thompson F-test - Excellent for time-varying line noise - Well-tested in neuroscience community - Can automatically detect line noise frequencies

Disadvantages: - Requires MATLAB license (commercial) - Requires CleanLine plugin installation in EEGLAB - Slower due to Python-MATLAB communication - Most computationally intensive method - Higher memory usage

Best for: Users with MATLAB + EEGLAB setup who need the highest quality adaptive filtering.

Installation Requirements: 1. MATLAB (license required) 2. EEGLAB installed 3. CleanLine plugin installed in EEGLAB 4. MATLAB Engine for Python 5. EEGLAB in MATLAB path

4. MATLAB: IIR Notch Filter

Type: Infinite Impulse Response notch filter Speed: ⚡⚡ Medium Quality: ⭐⭐⭐⭐ Very good

Native MATLAB implementation using iirnotch and filtfilt functions.

Advantages: - Native MATLAB implementation - Very narrow-band filtering possible - Well documented - Compatible with MATLAB workflows

Disadvantages: - Requires MATLAB license (commercial) - Slower due to Python-MATLAB communication - Not adaptive

Best for: Users with existing MATLAB license who prefer MATLAB-native implementations.

5. Octave: IIR Notch Filter (Free)

Type: Infinite Impulse Response notch filter via GNU Octave Speed: ⚡ Slow Quality: ⭐⭐⭐⭐ Very good

MATLAB-compatible filtering using free GNU Octave.

Advantages: - Free and Open Source - MATLAB-compatible syntax - Similar results to MATLAB - No license costs

Disadvantages: - Octave and oct2py must be installed separately - Slower due to Python-Octave communication - ~95% MATLAB compatible (minor differences possible) - Not adaptive

Best for: Users without MATLAB license who want MATLAB-like processing.

Interface Components

Method Display

  • Shows currently selected method and region
  • Example: Method: ⭐ MNE Spectrum Fit | 60 Hz (US)
  • Clickable - opens detailed method information dialog
  • Updates automatically when settings change

Files Counter

  • Shows current progress as "Files: processed/total"
  • Updates in real-time during processing
  • Example: "Files: 3/10"

Progress Bar

  • Visual progress indicator showing percentage complete
  • Only appears during processing
  • Automatically hidden after completion

Action Buttons

  • "Methods Info" - Opens detailed comparison of all methods
  • "Remove Noise" - Starts processing with selected method

Process Flow

  1. Initialization
  2. System checks for input files from Grid Association
  3. Verifies selected method availability (MATLAB/Octave)
  4. Updates method display

  5. Enables/disables action button based on validation

  6. Processing Start

  7. User clicks "Remove Noise" button
  8. Button shows loading animation
  9. Progress bar appears
  10. Files counter resets

  11. Output directory is created

  12. File Processing

  13. Files are processed sequentially
  14. Progress bar updates after each file
  15. Files counter increments

  16. Results saved to line_noise_cleaned/ folder

  17. Completion

  18. Loading animation stops
  19. Progress bar disappears
  20. Final file count displayed
  21. Step marked as complete
  22. Success message shown

Configuration

Method Selection

Configure the line noise removal method in:

Settings > Preferences > Line Noise Removal > Line Noise Removal Method

Available options (depending on installed software): - ⚡ MNE-Python: Notch Filter (FIR) - Fast - ⭐ MNE-Python: Spectrum Fit (Adaptive) - Recommended - 🏆 MATLAB: CleanLine (EEGLAB Plugin) - Gold Standard - 🔬 MATLAB: IIR Notch Filter - 🐙 Octave: IIR Notch Filter (Free)

Region Selection

Configure your powerline frequency region in:

Settings > Preferences > Line Noise Removal > Power Line Frequency Region

Options: - 🇺🇸 USA/North America (60 Hz) - Removes 60, 120, 180, 240 Hz - 🇪🇺 Europe/Asia (50 Hz) - Removes 50, 100, 150, 200 Hz

File Handling

Input files: - Located in the associated_grids/ directory - Must be in .mat format - Must contain valid EMG channel data

Output files: - Saved in the line_noise_cleaned/ directory - Maintain original file names - Contain the same structure with cleaned signal data

Method Comparison

Method Speed Quality Cost Adaptive Dependencies
MNE Notch ⚡⚡⚡ ⭐⭐⭐ Free ✗ MNE-Python
MNE Spectrum Fit ⚡⚡ ⭐⭐⭐⭐⭐ Free ✓ MNE-Python
MATLAB CleanLine ⚡ ⭐⭐⭐⭐⭐ MATLAB ✓✓✓ MATLAB + EEGLAB
MATLAB IIR ⚡⚡ ⭐⭐⭐⭐ MATLAB ✗ MATLAB
Octave IIR ⚡ ⭐⭐⭐⭐ Free ✗ Octave

Error Handling

Common Issues

  1. Method Not Available
  2. Warning: "MATLAB is not available. Please install MATLAB Engine..."

  3. Solution: Install required software or select a different method

  4. No Input Files

  5. Warning: "No files available for processing"

  6. Solution: Ensure Grid Association step was completed successfully

  7. Processing Errors

  8. Error displayed in step status
  9. Check application logs for details
  10. Verify file integrity

MATLAB/CleanLine Specific Issues

  1. EEGLAB Not Found
  2. Error: "EEGLAB might not be on MATLAB path"

  3. Solution: Add EEGLAB to MATLAB path in startup.m

  4. CleanLine Not Installed

  5. Error: "CleanLine execution failed"

  6. Solution: Install CleanLine plugin in EEGLAB

  7. MATLAB Engine Issues

  8. Error: "Failed to start MATLAB engine"
  9. Solution: Reinstall MATLAB Engine for Python

Best Practices

Method Selection

  1. For most users: Use MNE Spectrum Fit (default)
  2. Excellent quality
  3. Free and fast

  4. No additional software needed

  5. For MATLAB users with EEGLAB: Use CleanLine

  6. Gold standard quality
  7. Best for time-varying line noise

  8. Requires MATLAB setup

  9. For quick testing: Use MNE Notch Filter

  10. Very fast
  11. Good enough for initial exploration

Before Processing

  1. Verify method availability in settings
  2. Check input file availability
  3. Ensure sufficient disk space
  4. Select correct powerline frequency region

During Processing

  1. Allow each file to process completely
  2. Monitor progress bar and file counter
  3. Don't close the application during processing
  4. Watch for error messages in step status

After Processing

  1. Verify output files in line_noise_cleaned/ folder
  2. Check file counts match expected totals
  3. Visually inspect a few cleaned signals (optional)
  4. Proceed to next step only after completion

Technical Details

MNE Spectrum Fit Algorithm

The spectrum fit method uses a sophisticated approach: 1. Segmentation: Signal divided into overlapping windows 2. Spectral Analysis: FFT applied to each window 3. Sinusoid Fitting: Sinusoidal curves fitted to interference frequencies 4. Subtraction: Estimated interference subtracted from signal 5. Smoothing: Transitions between windows smoothed

MATLAB CleanLine Algorithm

CleanLine uses multi-taper spectral analysis: 1. Multi-Taper Estimation: Uses Slepian sequences (DPSS) for robust spectral estimation 2. Statistical Testing: Thompson F-statistic tests significance of line noise 3. Adaptive Fitting: Fits sinusoids with time-varying amplitude and phase 4. Regression: Least-squares regression estimates interference parameters 5. Removal: Subtracts estimated interference while preserving signal

Key CleanLine Parameters: - Window size: 4 seconds (default) - Window overlap: 50% (default) - Bandwidth: 2 Hz per frequency - Significance level: p-value threshold - Smoothing factor: 100

Data Format

All methods: - Preserve original data structure - Maintain sampling frequency - Keep all metadata - Only modify signal data array

Installation Guides

MNE-Python (Already Installed)

pip install mne>=1.0.0

MATLAB Engine for Python

Installation Assistant (Recommended): 1. Go to Settings → Preferences → Line Noise Removal 2. Scroll to MATLAB Engine for Python section 3. Click Show Installation Instructions button 4. Follow the displayed instructions (copy-paste ready commands) 5. Click "Copy Command" to copy the installation command 6. Run the command in MATLAB or Terminal/CMD 7. Restart application after installation

Benefits: - Automatically detects your MATLAB installation path - Provides correct commands for your system - Works with all Python environments - Simple copy-paste workflow

Manual Installation (if you prefer):

Option 1 - In MATLAB:

cd(fullfile(matlabroot,'extern','engines','python'))
system('python setup.py install')

Option 2 - In Terminal/CMD:

cd <matlabroot>/extern/engines/python
python setup.py install

MATLAB CleanLine (EEGLAB Plugin)

  1. Install MATLAB (license required)
  2. Install EEGLAB from sccn.ucsd.edu
  3. In EEGLAB: File → Manage EEGLAB extensions → CleanLine
  4. Install MATLAB Engine (see above - use automatic installation)
  5. Add EEGLAB to MATLAB path (in startup.m or manually)

Octave

  1. Install Octave from octave.org
  2. Install Python package:
pip install oct2py>=5.0.0

Next Steps

After line noise removal is complete: 1. Verify the number of processed files matches your input 2. Check output directory for cleaned files 3. Proceed to Crop to Region of Interest

Troubleshooting

If you encounter issues:

  1. Check Installation Status
  2. Open Settings → Preferences → Line Noise Removal

  3. Verify installation status indicators

  4. Method-Specific Issues

  5. MATLAB/CleanLine: Check EEGLAB and CleanLine installation
  6. Octave: Verify Octave installation with octave --version

  7. All methods: Check application logs

  8. Performance Issues

  9. CleanLine and Octave methods are slower - this is normal
  10. For faster processing, use MNE Spectrum Fit

  11. Ensure sufficient RAM for large files

  12. Quality Issues

  13. Verify correct region selected (50 Hz vs 60 Hz)
  14. Try different methods to compare results
  15. Check if line noise is actually present (use Methods Info for guidance)

References