github francescopace/espectre 1.2.0
v1.2.0 - Simplified Architecture & MVS Segmentation

latest releases: 2.3.0, 2.2.0, 2.1.0...
one month ago

🛜 ESPectre 👻 - v1.2.0

🏗️ Major Refactoring - Simplified Architecture

Architectural simplification: Focus on segmentation-only approach

This release removes the complex calibration/detection system in favor of a simpler, more maintainable segmentation-only architecture.

Key Changes

  • Removed modules (6 files, ~2,845 lines removed, -60% complexity):

    • calibration.c/h - Complex calibration system
    • detection_engine.c/h - Multi-state detection engine
    • statistics.c/h - Statistics buffer and analysis
  • Simplified core files:

    • espectre.c: 900 → 557 lines (-38%)
    • mqtt_commands.c: 966 → 604 lines (-37%)
  • Cleaner MQTT API: 10 commands instead of 18

✨ New Features

Moving Variance Segmentation (MVS)

  • Real-time motion segment detection and analysis
  • Adaptive threshold based on moving variance
  • Circular buffer maintains up to 10 recent segments
  • Foundation for advanced motion pattern recognition

Amplitude-Based Skewness & Kurtosis

  • +151% separation improvement
  • Skewness: 2.91x separation, 82.3% accuracy, 0% false positives
  • Kurtosis: 2.47x separation (+79% improvement)

Improved Traffic Generator

  • ICMP ping-based (replaced UDP broadcast)
  • Reliable CSI packet generation on every ping reply
  • Auto-discovery of WiFi gateway
  • Configurable rate: 1-50 pps

CSI Subcarrier Optimization

  • Now captures ALL 64 subcarriers (+23% spatial information)
  • Better movement detection accuracy

Local Segmentation Test Script

  • Python tool for rapid parameter tuning without device flashing
  • Parameter optimization with grid search
  • Interactive visualization

🔧 Changes

  • Temporal Features: Expanded from 8 to 10 features
  • Modified Fisher Criterion: Improved feature selection algorithm
  • Removed: Adaptive normalizer filter (code simplification)

📊 Benefits

  • ✅ Simpler codebase (~2,845 lines removed)
  • ✅ Easier to understand and maintain
  • ✅ Focus on proven MVS algorithm
  • ✅ Optional feature extraction (performance optimization)
  • ✅ Better detection accuracy

📚 Documentation

  • Updated README with changelog badge and links
  • Complete CHANGELOG.md with detailed changes
  • Improved badge interactivity

Full Changelog: https://github.com/francescopace/espectre/blob/main/CHANGELOG.md

Don't miss a new espectre release

NewReleases is sending notifications on new releases.