HUMAN ACTIVITY RECOGNITION USING MACHINE LEARNING

Abstract

This study utilizes the Human Activity Recognition UP Fall dataset, consisting of data from 11 different physical activities such as Walking, Cycling, Jogging, Running, Lying Down, Climbing Stairs, Sitting, Standing, Jump front and back, Frontal elevation of Arms, none are performed by 10 young adults age between 18-24 was wearing 2 inertial measurement units using two sensors namely accelerometer and gyroscope. The increasing popularity of wearable devices like smart watches, smart phones, and wristbands has generated a need to analyze user patterns and activity relationships. However, the existing algorithm Logistic Regression has lesser accuracy and often fail in complex environment dealing with the large datasets. This proposal presents the activity recognition, intensity estimation, and the development of algorithms for data processing, feature extraction, and classification. Leveraging this enriched dataset, machine learning techniques are applied to accurately predict the specific activity of a user is engaged in. The approaches used to implement the proposal are the K-Nearest Neighbors (KNN) algorithm, Random Forest algorithm, Grid Search CV for best n_neighbours and Robust Scaler to standardize the features.