Bear Interaction Activity Classification and IMU Signal Processing

This repository contains the codebase for collecting, processing, visualizing, and classifying IMU sensor signals from a bear interaction enrichment activity system.

Project Scope

• IMU data collection on M5StickC Plus2 (m5stick/)
• Signal preprocessing and ELAN export utilities (Data collection/)
• Plotting and activity classification scripts (Plotting/)
• Activity classes used in experiments: poke, sway, tilt, spin

Repository Structure

• m5stick/: Firmware sketches for on-device IMU capture and retrieval
• Data collection/: Raw/processed datasets, behavior sessions, ELAN formatting scripts, and graph outputs
• Plotting/: Visualization scripts, confusion matrix generation, and decision-tree based activity classification

Quick Start

1. Install Python dependencies:
   • pandas
   • numpy
   • matplotlib
   • scikit-learn
2. Run plotting/classification scripts from the Plotting/ directory, for example:
   • python plot.py
   • python confusion-matrix.py
   • python decision-tress.py

Research Images

Sensor Signal Visualization

Model Result Visualization

Behavior Capture Example

Notes

• Generated and raw files are kept in their original folders for reproducibility.
• This repo is focused on bear interaction activity classification and IMU signal processing artifacts.