Representation Learning from Time Labelled Heterogeneous Data for Mobile Crowdsensing

Figure 1

The architecture of the deep belief network for mobile crowdsensing. It contains four layers. The input data is a vector that is collected and integrated from accelerometer, gyroscope, magnetometer, compass, and time sequence.