In this project, I used Python and various machine learning models to detect human activity from CSI (Channel State Information) packet data collected via Wi-Fi router antennas.
Each activity is represented by an integer label:
- 1: Lie down
- 2: Fall
- 3: Walk
- 4: Pick up
- 5: Run
- 6: Sit down
- 7: Stand up
To recognize different human movements based on signal data and evaluate classifier accuracy and performance.
- K-Nearest Neighbors (KNN) using different values of
k(1, 2, 5, 10, 15) - Support Vector Machine (SVM) with default hyperparameters
- Random Forest Classifier with 5, 10, and 20 estimators
- Accuracy scores were calculated for all models.
- Normalized Confusion Matrices were plotted to visualize classification performance and misclassifications.
- Models were compared based on both accuracy and computational complexity.
- Shape of training data:
(3977, 250, 90) - Shape of test data:
(500, 250, 90) - Labels: scalar values for each activity
🛠️ Data reshaping was necessary before training, as machine learning models require 1D input features.