Personal Health Indicators by Deep Learning of Smart Phone Sensor Data
Honggui Li, Maria Trocan · 2017
This paper attempts to build the correlation between smartphone sensor data and personal health by deep learning. Firstly, the data from smartphone sensors is classified by deep Stacked Auto Encoders (SAE). For example, accelerometer data can be categorized by user motion states, such as standing, walking, and running. Secondly, the quantitative relationship between divided sensor data and human health is established. For instance, walking and running are helpful for human health. Finally, simulation experiment is conducted to evaluate the proposed method. In the situation of supervised feature extraction, a maximum 5 percent gain on overall classification accuracy on three open databases has been obtained compared with the state of the art methods. It is also testified by the experimental results that smartphone sensor data can reflect the healthy information of the user to some extent.