Abstract:
Human activity recognition (HAR) is an important part of people's daily lives because it allows
them to get high-level information about human actions from raw sensor input data. HAR is the
difficulty of characterizing daily human activity using data acquired from smartphone sensors in a
single statement. The accelerometer and gyroscope collect data on a continual basis, and these data
are crucial in forecasting human behaviors like walking or standing. On this subject, there are
several databases and continuing study. The COVID-19 pandemic is the world's most pressing
problem now, thus it's critical to follow or record the everyday actions of those who live alone or
isolate themselves. As a result, human activity recognition is crucial in the medical profession.
Human Activity Identification (HAR) is used in a variety of applications, including eldercare,
medical, sport activity monitoring, surveillance, emotion recognition, and training. Human
Activity Recognition is the subject of a lot of study and research. However, in most of the paper,
there are just two or three models. What we know is that the more models we evaluate with more
data, the better model and accuracy we will find. Pre-processing of data, training and testing with
selected models, evaluation of outcomes (accuracy), and better model prediction for HAR are the
phases of the system model. I have taken the "Human Activity Recognition with Smartphones
Dataset (2019)" to apply machine learning methods. Kaggle was used to get this dataset. The
Human Activity Recognition database was created using recordings of 30 research participants
conducting activities of daily living (ADL) while wearing a smartphone with inertial sensors
attached to their waist. The datasets I took in this paper were separate for train and test, and the
data was taken to a smart device through a sensor as already mentioned in the dataset description.
If I compare between my proposed methods, then the Random Forest method works well in Human
Activity Recognition which is evidenced by the comparison table.
