Human Activity Recognition For Analysing Fitness Dataset Using A Fitness Tracker.

The rise in eminence in the research based on Human Activity Recognition (HAR) due to its extensive scope in its application such as monitoring activities of daily life, fitness, security, and many more areas, is no mystery. In this project, we will be concentrating on the application of HAR in mobile health. The demand for HAR has increased in recent times due to the rapidly growing elderly, obese, and overall unhealthy population. We will be dealing with a Smartphone-based HAR system as a fitness tracker that uses high-dimensional sensors to collect fitness datasets and infer a confusion matrix for HAR in different activities to analyze the best fit in this work and predict the most accurate activity. It is crucial to select the appropriate features to support our research. We have used the Random Forest feature selection and the extracted data was analyzed using different classification techniques along with the previously mentioned scheme to generate the most accurate results and find a method that can be used by developers in the fitness tracker industry. Our data is a collection of 30 participants within an age bracket of 19 - 48 years performing six activities namely, Walking, Walking Upstairs, Walking Downstairs, Sitting, Standing, and Laying while carrying a waist mounted smartphone with embedded wireless sensors. These activities recorded certain movements, acceleration, velocity, and heart rate which was compiled into readable data and worked upon.