Activity Recognition in a Home Environment Based on Silhouettes
Name
Cardo Kambla
Abstract
The amount and cost of human resources are big problems in healthcare. The availability of these resources is limited and therefore, it is not always possible to send help when someone needs assistance. A part of human resources goes into helping people at home who suffer from mental or physical disorders.
The purpose of this Bachelor’s thesis is to help health professionals by creating
a machine learning model, which can recognize human activity from a camera. The recorded material is extracted in a way, that the human and background details are removed to keep people’s identity anonymous, resulting in a picture, where there is a white silhouette on a black background.
The purpose was achieved by extracting stationary and non-stationary features for the machine learning model. K-Nearest neighbors and Random Forest classifier were used by the new features. The final results showed that it is possible to recognize activities from silhouettes however, the result depends a lot on the data and the use cases.
The purpose of this Bachelor’s thesis is to help health professionals by creating
a machine learning model, which can recognize human activity from a camera. The recorded material is extracted in a way, that the human and background details are removed to keep people’s identity anonymous, resulting in a picture, where there is a white silhouette on a black background.
The purpose was achieved by extracting stationary and non-stationary features for the machine learning model. K-Nearest neighbors and Random Forest classifier were used by the new features. The final results showed that it is possible to recognize activities from silhouettes however, the result depends a lot on the data and the use cases.
Graduation Thesis language
Estonian
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Meelis Kull
Defence year
2018