5.3.2. The Influence of Phase Shift Threshold *ϕ* on Experimental Results

An appropriate setting of the RFID phase threshold is the key to the algorithm for eliminating *π* phase jump and accurately estimating the object's moving radial velocity. In this paper, it is assumed that the phase difference of the same object in the adjacent time should not exceed the threshold. In this section, we set *ρ* = 0.1, *ξ* = 2, *w* = 25, *N* = 200. Table 4 shows the localization accuracy under different phase shift thresholds.

It can be seen from Table 4 that when *ϕ* is set too small (e.g., *ϕ* = 10◦ ), the algorithm will erroneously remove the originally correct phase information, resulting in the matching rate of only 81.6% and a localization error of 0.72 m. When *ϕ* is set too large (e.g., *ϕ* = 180◦ ), the algorithm cannot correctly handle the phase where the jump occurs, resulting in the localization error up to 0.83 m and matching rate is only 66.8%. Therefore, only by setting an appropriate phase shift threshold can we

better eliminate the effect of phase jumps while retaining normal phase information. We set *ϕ* = 90◦ , the matching rate can reach 90.2%, and the localization error is only 0.33 m, in order to obtain better experimental results.


**Table 4.** The influence of phase shift threshold *ϕ* on experimental results.

#### 5.3.3. The Influence of The Number of Particles *N* on The Experimental Results

In the traditional particle filter algorithm, the number of particles will have an effect on the results. In this paper, Pearson correlation coefficient is added to constrain the particles in the update stage of particle filter, so we carry out experiments to analyze the impact of the number of particles on the localization error and matching rate in this case. We used CPU with core i5-7300 HQ, 2.50 GHz, and 8 GB ram in the experiment, and set other parameters exactly the same as before. Table 5 lists the results.


**Table 5.** The influence of the number of particles *N* on the experimental results.

It can be seen in the above Table 5, a small *N* (such as *N* = 5) gives an increase in the localization error, since the small number of particles cannot effectively represent the probability density. The positioning accuracy gets improved when increasing the number of particles, similarly, performing filtering with the large number of particles also consumes more time. With a large *N*

1000 0.33 12.17

(such as *N* = 1000), we almost get the same localization results. For considering the accuracy and the time-consuming of the algorithm, we choose *N* = 200 in our experiment.
