Markerless Sensors for Physical Health Monitoring System Using ECG and GMM Feature Extraction

Physical activity monitoring using wireless sensors is highly demanded in medical fields such as heart attack monitoring, Parkinson's disease, sleep problems and disrupted circadian rhythms. For the productivity of these areas, many researchers are actively involved in physical activity monitoring. However, they still lack potency in term of recognition accuracy. This paper proposed a novel system that attempts to improve recognition accuracy by employing Electrocardiogram and Gaussian Mixture Model feature extraction strategy along with Ant Colony Optimization and Genetic Algorithm as a classifier. Unlike conventional results, our proposed model has shown a high validity results of 83.33% and 92.70% over the two-public benchmark IMSB and USC-HAD datasets, respectively. The proposed system can be used to monitor physical activities of human that can be helpful in daily life activities especially in medical field.