Recognition rate difference between real-time and offline human activity recognition

摘要

The appearance of the Internet of Things topic has a huge impact on several research fields including human activity recognition (HAR) where wearable sensors provide the raw information about the physical activity and functional ability of an observed person. Previous studies have shown that HAR can be seen as a general machine learning problem with a particular data pre-processing stage. In the last years, several researchers reached high recognition rates on public data sets or in laboratory environment but their solutions have not tested yet in real-life. Therefore, this paper investigates the efficiency of previously used machine learning strategies in real environment by an Android-base, self-learning HAR application which has been designed according to the latest HAR solutions. The result of this study shows a significant recognition rate difference between the “online” (real-time) and “offline” cases.