Method and system for multicarrier signal tracking based on deep learning and high precision positioning

A deep learning model training method of multicarrier signal using S-curve, comprising:

• generating original multicarrier baseband signals based on a protocol for targeted multicarrier signal;
• adding fixed delay resolution and main path delay with setting range to the original multicarrier baseband signals, and adding, based on the original multicarrier baseband signals, different signal-to-noise ratios, multipath delay, and multipath amplitude respectively;
• according to a protocol of collected multicarrier signal, coarse synchronizing the added multicarrier signals, and obtaining starting position of the added multicarrier signals;
• according to the protocol of multicarrier signal and obtained starting positions, demodulating the added multicarrier signals, and extracting an applied pilot frequency information of a received pilot signal;
• based on the early and late correlation of the received pilot signal and a reference pilot signal, obtaining S-curve corresponding to different signal parameters, by an Early Minus Late Power (EMLP) discriminator function;
• wherein the EMLP discriminator function is defined as:
• EMLP(Δτ)=|Re(Δτ)|2−|Rl(Δτ)|2=G2AS(Δτ,ξ)
• wherein Re (Δτ) and Rl (Δτ) are the early and late correlation outputs of the received pilot signal and the reference pilot signal, respectively, G is the number of pilot, and A is the gain of the received signal, and S(Δτ,ξ) is S-curve of the received signal;
• wherein S-curve feature data of each different main path delay label corresponding with different signal-to-noise ratios, multipath delay, and multipath amplitude parameters, using different main path delay as label;
• inputting a delay label and the S-curve feature data into deep learning model;
• wherein the deep learning model training will be completed to get training weight results when the loss function of the deep learning model training is stable.