Human Activity Recognition with Wearable Biomedical Sensors in Cyber Physical Systems

Human Activity Recognition has a wide range of applications such as remote patient monitoring, assisting disables and rehabilitation. This paper investigates the use of wearable bio-medical sensors to recognize human activities using supervised learning algorithms in cyber physical systems. We use five bio-medical sensors such as ECG, EMG, Respiration, Force sensitive resistor and a Tri-axial Accelerometer to collect the raw data. All the sensor data is collected in the real-world environment with three human subjects. The received raw data is preprocessed to extract the time domain features. The feature information is used for the training and testing the classifiers. Three classifiers k nearest neighbour (kNN), SVM using the linear kernel and SVM using Gaussian kernel are used for training and testing phases. The kNN classifier provides good accuracy of 99.86 %.