v2025.11

v2025.11

Changelog since 2025.10:

Added Features

1. Adaptive upstate prediction learning - Retrospective learning system that continuously improves slow oscillation upstate prediction accuracy by analyzing actual slow oscillation morphology throughout the session
2. Online re-referencing - Runtime channel re-referencing capability for improved signal quality without affecting stored raw data
3. Enhanced settings dialog - Comprehensive pre-session configuration GUI accessible via File → Session Settings menu
4. Post-session stimulation analysis - Grand average event-related potentials around audio stimulations for sound feedback mode validation
5. Phase analysis visualization - Polar plots of signal phase at predicted upstate times with Rayleigh statistics for phase-locking validation

Bug Fixes

1. Fix critical bug where predicted upstates were forward shifted due to growing onset array size
2. Correct audio feedback loop mode to properly respect refractory periods
3. Fix channel switching handler to work when real-time loop is not running
4. Align downstate detection with signal processing (index vs timestamp handling)

Improvements

1. Index-based plotting - Post-session analysis now uses sample indices instead of timestamps to eliminate cumulative timing errors
2. Enhanced accuracy metrics - Detection and prediction accuracy histograms with offline ground truth comparison
3. Channel switching logging - Reference and channel changes now logged to stimulation file with timestamps
4. Adaptive thresholding - Trough multiplication coefficient automatically adapts via rolling buffer averaging
5. Code organization - Comprehensive docstrings added to core signal processing functions
6. Removed duplicate logic - Consolidated sound feedback loop implementation

Technical Details

• Rolling buffer mechanism: 100-coefficient buffer with 3-sample averaging for stable adaptive learning
• 7-point validation criteria: Ensures upstate learning updates only on high-confidence slow oscillations
• Dual notch filtering: 50Hz and 60Hz line noise removal for international deployment
• Enhanced data outputs: Channel/reference switching events tracked in *_stim.txt files

What's Changed

This release introduces adaptive personalization that allows the system to learn individual-specific slow oscillation patterns during sessions, progressively improving stimulation timing accuracy. The online re-referencing feature enables real-time signal quality improvement without modifying stored data, while enhanced post-session analysis tools provide comprehensive validation of system performance.

Full Changelog: v2025.10...v2025.11

v2025.10

Changelog since 2025.09

Added features

1. Rewrite post-session report to be more straightforward and make use of optimized libraries (i.e. Pandas)
2. Added debugging modes: Offline Muse data simulation; Sound playback
3. Cover line noise across different regions (50 and 60 Hz)

Bug fixes

1. Fix shutdown sequence not completing

What's Changed

• 2025.10 by @SoftcodingForYou in #7

Full Changelog: v2025.09-BETA...v2025.10

v2025.09 Beta

What's Changed

• Implement usage of Muse devices by @SoftcodingForYou in #6

tl;dr

• Make code compatible with Muse devices
• Listener threads for background processes including data storage
• Small GUI for controlling signal management on the fly instead of triggering actions because of system-wide keystrokes

New Contributors

• @SoftcodingForYou made their first contribution in #6

Full Changelog: v2025.06...v2025.09-BETA

v2025.06

Fully functionning algorithm:

1. Study stimulation
2. Deep sleep slow wave cueing
3. Post-session report

Compatible boards:

• OpenBCI Cyton 8-Channel

EazzZyLearn-PRO-V1.0.0

First version of EazzZyLearn PRO in Helment HQ

All parameters to be adjusted inside parameters.py