AccelGuard is a MATLAB framework for real-time anomaly detection using accelerometer data. It analyzes three-axis readings to detect irregularities like free falls or machinery faults, with applications in wearable devices, industrial monitoring, robotics, and IoT systems, ensuring safety and operational efficiency.
- Real-Time Anomaly Detection: Quickly detects deviations from normal patterns in accelerometer data.
- Multi-Axis Analysis: Processes data from all three accelerometer axes (X, Y, Z) for comprehensive detection.
- Visual Outputs: Generates detailed graphs for identifying and interpreting anomalies.
- Customizable Thresholds: Allows users to tailor detection thresholds based on specific use cases.
- Free-Fall Detection: Identifies sudden drops or falls in wearable devices.
- Industrial Monitoring: Detects machinery vibrations or abnormal movements for predictive maintenance.
- Robotics: Monitors irregular robotic movements to ensure operational safety.
- Clone the repository:
git clone https://github.com/yourusername/AccelGuard.git
- Open MATLAB and navigate to the cloned directory.
- Ensure the required datasets are available in the working directory.
-
Load the datasets into MATLAB:
- Normal data (baseline).
- Anomalous data (known deviations).
- Testing data (evaluation dataset).
-
Run the main script:
run('AnomalyDetection.m') -
View results:
- Check the Command Window for intermediate outputs.
- Analyze the generated graph for detected anomalies.
AnomalyDetection.m: Main script for running the anomaly detection workflow.- Data Files: Placeholder for datasets (replace with your actual files).
- Results: Includes visualizations like anomaly graphs.
Below is an example of the output graph generated by AccelGuard:
The graph highlights detected anomalies as peaks or deviations from the baseline, making it easy to identify abnormal patterns.
Contributions are welcome! Feel free to fork this repository, make enhancements, and submit a pull request. Whether it’s improving the detection algorithm or adding new features, your input is valuable.
This project is licensed under the MIT License. See the LICENSE file for details.