Human Activity Recognition

Published on 18 May 2022. Written by Amitesh Kumar

Human Activity Recognition is the project meant for tracking the activities of humans in their daily life. This project is based on the pattern study and data filtration, which is obtained through machine learning. Machine learning will help to study the huge data variations of human activities. The project is highly useful in medical assistance, elder care, rehabilitation assistance, diabetes, and cognitive disorders domain.

Read more..

This project is especially used in smartphones where the patient can enter their health status as inputs, and then the application will give the results as healthy or unhealthy. The project aims to evaluate the feature selection and extract the relevant data required for the procedure. For the project, one needs to hold good knowledge about data patterns and python language.

Methods used

There are various data classifiers used in the project and also methods to evaluate the data. There is a total of six methods used for the application; some of them are the MLM (Minimum learning Machine), MLP (multi-layer perceptron), and Support Vector Machine (SVM). These methods are used as algorithms and used to solve the problem of the study of huge and complicated data.

Project Implementation

The first set is to feed the data into the methods. The data is usually of six parameters, which will be read by smartphones. The important parameters are walking, walking upstairs, walking downstairs, sitting, standing, and Lying. Based on these parameters, the application will work. The smartphone should be worn near the waist so that it could read the signals easily of body postures.

To extract the useful data and neglect the unnecessary information, Principal Components Analysis (PCA) is used. The approach is to catch the signals which are greater than the threshold signals, which means the normal signals. The technique will be known as Var-PCA after this method is applied. The threshold value should be accurate, and thus, classifiers will be used to reduce the lower variance.

The parameters for the classifiers that will be used chosen based on cross-validation. The inputs which are used should be accurate, and thus, the whole application becomes dependent on them. One should be careful using the methods as the coordination is important between them.

Results

The results displayed can be anything which you desire as 1-10, where 10 is the best being of health, or healthy/unhealthy, approach doctor/well-being, etc. This application will teach machine learning basics and the methods used within.

Kit required to develop Human Activity Recognition:

• Not Required

Technologies you will learn by working on Human Activity Recognition:

• Machine Learning

Any Questions?