Procedure

Below is the detailed procedure used for the CSI Fall Detection project. Click the image to enlarge.

Step Description Image
1 Prepare the data by merging the label and activity data files from the UT-HAR dataset. Merging data from the UT-HAR dataset Merging data from the UT-HAR dataset
2 Apply various filter techniques (Butterworth Low-Pass, Hampel, Discrete Wave Transform) to the data for denoising. Applying filter techniques for data denoising Applying filter techniques for data denoising
3 Implement the sliding window algorithm to segment fall events. Sliding window algorithm for fall event segmentation Sliding window algorithm for fall event segmentation
4 Extract features from the segmented data using the time, frequency, and time-frequency domains. Feature extraction from segmented data Feature extraction from segmented data
5 Input the data into a machine learning algorithm for training and prediction. Input data into machine learning algorithm Input data into machine learning algorithm
6 Run cross-validation to ensure good performance metrics. Cross-validation for performance metrics Cross-validation for performance metrics
7 Evaluate each model/algorithm on various metrics and determine the best-performing. Evaluation of models and algorithms Evaluation of models and algorithms