Robust First Path Detection Based on Two-Stage CNN for Unspecified UWB Pulse Shape

In this paper, we propose a technique by exploiting the two-stage convolutional neural network (CNN) to improve the ranging accuracy regardless of the pulse shape of the received signal in the ultra-wideband (UWB) system. UWB system could be degraded localization accuracy according to pulse shapes between transceiver. In particular, the capability of the receiver to detect a small signal level has been improved recently, and pulse shape compatibility will become even more important for UWB devices. There are no mandatory specifications or interoperability testing, as of now, so that future compatibility issues can exacerbate positioning accuracy. The proposed method shows that precise ranging even if an unspecified pulse shape-based signal is received.