Fall Detection
PRESENTED BY PRIAGUNG KHUSUMANEGARA
Timeseries Classification DTW
Data Accelerometer
Total Acceleration
a = 𝑋2 + 𝑌2 + 𝑍2
Fall Template
Activity Template
Classification Report
Accuracy Data Testing
Fall 1.00 20
Running 1.00 20
Upstairs 1.00 20
Walking 0.80 20
avg / total 0.95 80
Timeseries Classification: KNN & DTW
K-Nearest Neighbors algorithm (k-NN) is a non parametric method used for classification.
Algorithm:
1. Choose a value for k
2. Find the distance between unlabeled point and training points.
3. Find the k-nearest points to unlabeled points
4. Classify unlabeled point by a majority vote of its neighbors
Classification Report
Accuracy Data Testing
Fall 1.00 20
Running 1.00 20
Upstairs 1.00 20
Walking 1.00 20
avg / total 1.00 80

Fall detection