In order to improve indoor and outdoor positioning accuracy in various positioning environments (e.g. dense rban/urban/suburban/rural) in case of emergency location service, it is primarily required to enhance LTE signal based positioning echnology that provides signal coverage of 99% or more in Korea. Typical LTE Reference Signal Received Power (RSRP) fingerprint ositioning method includes the offline phase that generates an LTE RSRP fingerprint DB based on the collected data and the nline phase in which the generated DB is compared with the measured value of the terminal to calculate the position. To improve he positioning performance of the LTE RSRP fingerprint DB generated by the offline phase, it is required to collect its location and TE RSRP data densely, but it is expensive and laborious. As an alternative to this, further research is needed to estimate the LTE SRP fingerprint DB of non-collecting points based on the data collected roughly through war driving. In this paper, the following echnical feasibility on LTE RSRP fingerprint DB estimation for emergency location will be studied and analyzed. First, Analysis of LTE  SRP raw data on a specific LTE cell, which is collected through repeated war-driving in a specific test environment. Next, stimation of LTE RSRP fingerprint DB on a specific LTE cell using Gaussian process regression. Lastly LTE RSRP error analysis etween estimated and measured LTE RSRP fingerprint DB.

Keywords: LTE, RSRP, Gaussian process regression