AdaBoost Support Vector Machine Method for Human Activity Recognition

Abstract

Human activity recognition research is being implemented more and more as technology advances in computer vision. Many fields require activity recognition technology, such as theft detection or online exam cheating detection. One method that is widely used is AdaBoost. This study proposes the AdaBoost Support Vector Machine Method, a combination of the AdaBoost Method and the Support Vector Machine. The evaluation uses datasets for human activity recognition and compares them with other machine learning algorithms. The results obtained indicate that the proposed method has the highest performance compared to the tested algorithms. The highest accuracy in this study was 96.06%. It shows that SVM as an AdaBoost component is proven to be able to improve the performance of AdaBoost.

Authors 4

1. Intan Nurma Yulita

   Padjadjaran University

   Affiliation as printed

   Department of Computer Science, Universitas Padjadjaran, Bandung, Indonesia

   Research Center for Artificial Intelligence and Big Data, Universitas Padjadjaran, Bandung, Indonesia

2. Erick Paulus Aachen Leiden Institute Advanced Computer Sciences

   Leiden University · Padjadjaran University

   Affiliation as printed

   Department of Computer Science, Universitas Padjadjaran, Bandung, Indonesia

   Leiden Institute Advanced Computer Sciences, Leiden University, Leiden, Netherland

3. Asep Sholahuddin

   Padjadjaran University

   Affiliation as printed

   Department of Computer Science, Universitas Padjadjaran, Bandung, Indonesia

4. Dessy Novita

   Padjadjaran University

   Affiliation as printed

   Department of Computer Science, Universitas Padjadjaran, Bandung, Indonesia