BrainMaze: Signal Synchronization and Fusion Toolbox

A repository for the signal synchronization and fusion project.

Project status

This is the initial repo skeleton. It contains a Software Requirements Specification (SRS) describing the goals, functional and non-functional requirements, and a proposed Python API for aligning and merging electrophysiology time-series from two MEF recordings.

High-level description

brainmaze_sigfusion is a Python utility intended to:

• Compute a time-varying synchronization map between two MEF3 recordings with different duration and sampling rates.
• Allow saving/loading of the computed alignment map to avoid re-computation.
• Provide a unified reader that presents a single, temporally-aligned data stream to downstream analysis code while preferring a designated source when overlapping data exists.

Key design points:

• Two-stage alignment: coarse global alignment (envelope-based) and fine, chunk-based local alignment to handle non-linear clock drift and periodic artifacts.
• Memory- and performance-conscious processing (streaming, downsampling, and vectorized operations where possible).
• Sampling-rate independent: the system will handle differing input sample rates and present data at a chosen target or preferred-source rate.

Where to find the Software Requirements

The full Software Requirements Specification (SRS) is in:

software_requirements/sw_req.md

Open that file for the detailed requirements and a minimal proposed Python API.

Repository layout (current)

• documents/
  • sw_req.md — Software Requirements Specification (SRS).
• docs/
  • testing.md — Testing and pytest implementation guide.

Testing

Development dependencies are declared in pyproject.toml under the dev optional-dependencies group. To set up a development environment and run tests, see docs/testing.md.

Recommended next steps

1. Implement a minimal SynchronizationMap class (save/load) and add unit tests.
2. Implement BrainmazeSync.compute_alignment using envelope downsampling for coarse alignment and a chunk-based refinement loop. Use mef_tools for lazy reading of MEF data.
3. Implement UnifiedMefReader.read_data with preference/fallback logic and stitching across sources.
4. Add a small CLI or notebook demonstrating alignment on sample data.

Contact / Notes

If you want me to scaffold the initial implementation or add unit tests / CI, tell me which piece to start with and I'll create a small, runnable prototype and tests.