Human activity recognition (HAR) recently has garnered a lot of attention in recent years especially in wearable smart devices due to its high demand in various application domains. Smart-devices are nowadays a perfect way of collecting personalized data from users. The data collecting from sensors in smart devices can be used for many purposes like health care, lifelogging or fitness. In this research, we propose an approach to have efficient and effective activity recognition with real-life data coming from smart devices. The proposed approach is based on Deep Convolutional Neural Network learning and is combined with an active learning approach which adds personalized recognition based on personalized data to the model. Experiments show the proposed model has the perfect efficiency with the state-of-the-art approach and with the personalized data and the personalized recognition for walking while holding a phone in different positions, and for all positions of holding phone and laying. Beside locomotion activities, more actions like sweeping, vacuuming have been analyzed with promising results.

Additional Metadata
Keywords Deep learning, Human Activity Recognition, Wearable devices sensor data
Persistent URL dx.doi.org/10.1109/HealthCom.2018.8531150
Conference 20th IEEE International Conference on e-Health Networking, Applications and Services, Healthcom 2018
Citation
Fekri, M. (Maryam), & Shafiq, M.O. (2018). Deep convolutional neural network learning for activity recognition using real-life sensor's data in smart devices. In 2018 IEEE 20th International Conference on e-Health Networking, Applications and Services, Healthcom 2018. doi:10.1109/HealthCom.2018.8531150