Intelligent System for Human Context Recognition
Sumair Aziz, Muhammad Umar Khan, Ahmad Zahoor, Syed Zohaib Hassan Naqvi · 2020 International Conference on Computing and Information Technology (ICCIT-1441) · 2020
Human-Centered computing technique aims to understand human behavior and can serve in fitness observing aging care, and numerous different spaces. Recognizing Human Context at their normal routine life is a very challenging task as behaviors vary from person to person. The current methods available for the detection of human activities are less accurate, inefficient, and are not effective to be used for recognition purposes. This study focuses on the recognition of human activities by using MPU-6050 signals, which is an efficient, cheap and relatively new method as compared to the existing practices. A total of 37 subjects were involved in this work for data acquisition and asked them to do their activities as usual while carrying the MPU-6050 sensor on their chest. Empirical Mode Decomposition (EMD) was employed for the de-noising the signals. Extensive experimentation resulted in the selection of two features, which were giving the best intra-class difference through Support Vector Machines (SVM). An accuracy of 100% was achieved using Quadratic kernel for SVM.