Page 189 - Kaleidoscope Academic Conference Proceedings 2021

Connecting physical and virtual worlds

Assuming that the CIR on subcarrier k at the moment t is The power distribution of multiple time-delay propagation
h(k, t) in the 5G NR system, where the value range of paths is obtained by Equation (6), and the channel peak
subcarrier k is k = 0,1,2, . . . ,6N RB − 1. Using the Inverse response ratio in the time domain can be obtained by
Fast Fourier Transform (IFFT), we can obtain the CIR Equation (8):
h(τ, t) in the time domain as Equation (4).
max�P(i,t)�
γ(t) = N P −1 （8）
h(τ, t) = IFFT(h(k, t), N IFFT ) （4） ∑ i=0 P(i,t)
Where N IFFT is the IFFT-point. The CIR energy in the time The peak path can be considered as the main path, where
domain is calculated by Equation (5). the power ratio of the main path signal to the received
signal is calculated. For the LOS propagation environment,
P(τ, t) = |h(τ, t)| （5） the channel power peak response ratio will be relatively
2
large, close to 1. For the NLOS propagation environment,
Assuming that the transmitted signal has one antenna port, the channel power peak response ratio will be relatively
and then P(τ, t) has only one peak. Then we pick out small, and the richer the multipath, the smaller the value.
several energy points near the peak as Equation (6)
2.4 Channel time correlation characteristic
P(i, t) = Ϝ[P(τ, t)], i = 0,1,2, . . . , N − 1 （6） The time-varying characteristics of the channel are caused
P
by the relative motion between the mobile station and the
where Ϝ[ ∙ ] represents the windowing operation around the base station. The Doppler spread and the correlation time
peak. How many energy points before and after power peak are the two parameters of the channel frequency dispersion
in the time domain should we take? In a 4G or 5G system, and the time-varying characteristic. The Doppler spread is
the Cyclic Prefix (CP) is added to alleviate the Inter- used as a channel feature to distinguish different scenarios
Symbol Interference (ISI) caused by the multipath delay. in [14]. In this paper, we use CIR correlation characteristics
Therefore, the multipath delay spread is mostly distributed in the time domain to distinguish different speeds of mobile
within the CP length. So N is not less than the CP length. station.
P
In order to combine with a neural network, Equation (6) Assuming that h(k, s, t) is the CIR on subcarrier k and pilot
needs to be normalized as Equation (7). symbol s at the moment t. Here k is the subcarrier index,
− 1. Here s is the symbol index,
P(i,t) and k = 0,1,2, . . . ,6N RB
P �(i, t) = i = 0,1,2, . . . , N − 1 （7） and s = 0,1. Then the time correlation value of CIR on two
P
max�P(i,t)�
pilot symbols can be calculated by Equation (9).
Then P �(i, t) can characterize the energy distribution of
1
multiple delay propagation paths, and it is also the channel β(t) = 6N RB −1 ∑ 6N RB −1 �conj�h(k, s = 0, t)� ∙ h(k, s = 1, t)� (9)
k=0
characteristic in the time domain in this paper. The larger
the number of IFFT points, the higher the resolution of The CIR power on the two pilot symbols is obtained by
multiple delay paths, and the smaller the number of IFFT Equation (10) and Equation (11) respectively.
points, the lower the resolution of multiple delay paths.
1 6N RB −1 |h(k, s = 0, t)| （10）
2
2.3 Channel power peak response ratio in time P (t) = 6N RB −1 ∑ k=0
1
domain
1 6N RB −1
2
P (t) = ∑ k=0 |h(k, s = 1, t)| （11）
2
When there are obstacles between the transmitter and 6N RB −1
receiver, the wireless signal can only reach through the In order to combine with a neural network, the time
scattering, diffraction and reflection paths. This kind of correlation value of the CIR needs to be normalized, as in
propagation environment is called NLOS. But if there is a Equation (12).
direct path, it is called LOS. The signal strength value is
used to distinguish LOS/NLOS in [6] and [7]. In [10], some 2
parameters such as Rician-K factor, kurtosis of the received β norm (t) = real[β(t)] （12）
power, skewness of the received power and angle P 1 ∙P 2
information are used as the basis for distinguishing The correlation value can characterize the channel change
LOS/NLOS. Here the high-order characteristics of the caused by speed, and the phase of the correlation value can
signal power are needed to calculate for the kurtosis and the also characterize the Doppler frequency offset, which can
skewness, and which is extremely challenging for hardware be obtained by Equation (13).
implementation. In this paper, the channel peak response
ratio in the time domain is used as the basis for angle�β(t)�
distinguishing LOS/NLOS. φ(t) = π （13）

– 127 –