electronics

Article
A Light-Weighted Machine Learning Approach to Channel
Estimation for New-Radio Systems
Hyun Woo Lee and Sang Won Choi *

Department of Electronic Engineering, Kyonggi University, Suwon 16227, Republic of Korea;

[email protected]
* Correspondence: [email protected]

Abstract: In this paper, we provide a light-weighted Machine Learning (ML) approach to channel
estimation for New-Radio (NR) systems. Specifically, based on the equivalence between the Channel
Impulse Response (CIR) in the time domain and its corresponding Channel Frequency Response
(CFR) in the frequency domain, the light-weighted ML model for the channel estimation is shown to
be established in comparison to the existing ML-based channel estimator. Furthermore, for practical
use, the quantized weights for the light-weighted ML-based estimator are shown to be feasible
without significant performance degradation in the sense of mean square error (MSE), which shows
the effectiveness of the proposed approach from the perspective of memory overhead. Consequently,
we show that there exists Signal to Noise Ratio (SNR) gain in comparison with the existing ML-based
estimator, which is validated by numerical results considering the Sounding Reference Signal (SRS)
for NR in the 3rd Generation Partnership Project (3GPP).

Keywords: channel estimation; Channel Impulse Response (CIR); Channel Frequency Response
(CFR); Machine Learning (ML); New-Radio (NR); quantization; Sounding Reference Signal (SRS); 3rd
Generation Partnership Project (3GPP)

1. Introduction
Orthogonal Frequency Division Multiple Access (OFDMA), an MA scheme for New-
Citation: Lee, H.W.; Choi, S.W. A Radio (NR), has gained widespread adoption in wireless communication systems owing
Light-Weighted Machine Learning to its resilience against frequency-selective fading channels. The receiver in wireless
Approach to Channel Estimation for communication receives a distorted signal due to delay spread, which is the multipath
New-Radio Systems. Electronics 2023, effect. Therefore, the channel must be estimated to compensate for the distorted signal.
12, 4740. https://doi.org/10.3390/ In general, channels are estimated using pilot signals known to both the transmitter and
electronics12234740 receiver. Since the pilot signals used vary depending on the user scenario, channels are
Academic Editor: Xiongwen Zhao
estimated by generating signals such as the Demodulation Reference Signal (DMRS) [1]
and Sounding RS (SRS) [2–5].
Received: 30 October 2023 Existing channel estimation methods usually estimate channels using the Least Square
Revised: 17 November 2023 (LS) and Minimum Mean Square Error (MMSE). The LS method estimates the channel
Accepted: 20 November 2023 by assuming that the channel is deterministic. Specifically, the LS uses only pilot signals
Published: 22 November 2023
except for the channel’s statistical information and the computational complexity is quite
low. However, the statistical information of the channel is not used, so the performance
is relatively low from the perspective of MSE. The MMSE method utilizes the statistical
Copyright: © 2023 by the authors.
information of the channel to exhibit optimal performance from the MSE perspective.
Licensee MDPI, Basel, Switzerland. However, there is a disadvantage that optimal performance can be achieved only when the
This article is an open access article statistical information of the channel is properly known, and the computational complexity
distributed under the terms and is relatively high compared to the LS method. It is necessary to study an improved
conditions of the Creative Commons channel estimation method to compensate for the problems of these existing channel
Attribution (CC BY) license (https:// estimation methods.
creativecommons.org/licenses/by/ On the other hand, a variety of Machine Learning (ML)–based estimation methods
4.0/). have been studied to compensate for the shortcomings of existing estimation methods.

Electronics 2023, 12, 4740. https://doi.org/10.3390/electronics12234740 https://www.mdpi.com/journal/electronics

Electronics 2023, 12, 4740 2 of 15

Table 1 depicts representative studies [6–16] conducted for achieving comparable perfor-
mance to the MMSE method using ML without the statistical information of the channel in
comparison with main contributions from this study. It is noteworthy to design the channel
estimation method using the black-box characteristic of ML. However, ML-based Channel
Frequency Response (CFR) estimation methods have the disadvantage of high complexity
because inputs and outputs are proportional to the length of the pilot signal, and the num-
ber of hidden layers is typically large. In the case of ML-based Channel Impulse Response
(CIR) estimation, it is complicated to use because the channel is estimated at each time
instance of the received signal in the time domain, or the received signal in the frequency
domain is estimated through multiple MLs. Furthermore, memory overhead is likely to
occur as memory requirements to store ML increase in proportion to the number of hidden
layers and nodes.

Table 1. A concise comparison of our work with the existing ML-based channel estimations.

Perspective Contents
· Machine learning using received signal [6]
· Machine learning using estimated channels [7–13]
Input data type in ML
· Machine learning using transmitted and received signals [14]
· Machine learning using transmitted signal, received signal, and estimated channels [15,16]
· ML to estimate instantaneous channels from one symbol [6–9,16]
· ML to estimate the channel for the current symbol using the channel estimated
Design the number of usage symbols
from the previous symbol [10,11,14,15]
according to channel types
· ML to perform interpolation to estimate the channel of the slot where pilot symbols are
located not only that of the pilot symbol [12,13]
· Estimation of Channel Impulse Response in time domain [14,16]
Channel estimation by domain types
· Estimation of Channel Frequency Response in frequency domain [6–13,15]
· Low complexity and low MSE performance due to the use of simple ML structure
(with or without one hidden layer) [8,16]
Complexity and MSE performance
· High complexity and high MSE performance from leveraging complicated ML structure
(MNN or CNN, LSTM, and so on) [6,7,9–15]
· Estimation of channel impulse response using received signal in the frequency domain
Our work · Single slot design for channel estimation with a one-time symbol
· Low complexity and high MSE performance with the simple ML structure

In this paper, we propose a memory-saving ML-based channel estimation for estimat-

ing CIR with low complexity. The ML model adopts Deep Neural Networks (DNNs) to
design as simple as possible with one hidden layer, and it uses ML’s black box characteris-
tics to input received signals in the frequency domain and output CIR in the time domain.
The weights of the trained ML can be stored by converting them from floating point 32 to
integral point 8 using the weight quantization method [17,18].
The rest of this paper is organized as follows. The system model for the process before
using the channel estimation method is introduced in Section 2. The existing channel
estimation and the simplest ML-based CFR estimation are presented in Section 3. The
proposed method in a general environment is presented in Section 4. A simulation analysis
of the proposed method and simulation analyses based on additional perspectives are
presented in Section 5. Our main conclusion is given in Section 6.
Notations: Vectors and matrices are written in bold letters. E[X] is the expectation of the
random vector X. (·) T and (·)† represent transpose and Hermitian transpose, respectively.
||x|| and ⊗ are the norm of the vector x and convolution, respectively. b x c is the nearest
integer to x, and Re(x) and Im(x) are real and imaginary values of the vector x, respectively.

2. System Model
This paper considers the comb-type pilot arrangement-based OFDM system [19–22] with
KTC intervals for subcarriers of length Nsc , as shown in Figure 1. Figure 2 shows a typical block
Electronics 2023, 12, 4740 3 of 15

diagram of the OFDM system with the pilot signal assisted. S̄ = [S0 , . . . , S N −1 ] T is an N × 1
pilot signal vector known by both transmitters and receivers, and X̄ = [ X0 , . . . , X Nsc −1 ] T is
an Nsc × 1 transmission signal vector in a frequency domain where pilot signal S̄ is placed
at regular intervals in subcarriers, and symbols without pilot signals are set to 0. After pilot
insertion, X̄ adds the Nzp × 1 Zero Padding (ZP) vector 0 Nzp to both sides to expand to
the Nfft × 1 vector, and as shown in Equation (1), X̄ with ZP added is sent to an Nfft -point
Inverse Fast Fourier Transform (IFFT), which is modulated into x in the time domain as

x = F− 1 T
N [ 0 Nzp , X̄, 0 Nzp ] , (1)
fft

where F− 1 1
N is an Nfft -point IFFT matrix and Nzp = 2 ( Nfft − Nsc ).
fft

Figure 1. Comb-type pilot arrangement for OFDM (KTC = 4).

Figure 2. Block diagram in OFDM for channel estimation.

The transmission signal x places a cyclic-prefix (CP) of length Ncp at the front of the
signal to prevent Inter-Symbol Interference (ISI), and then a ( Nfft + Ncp ) × 1 vector xcp is
transmitted to the receiver. Assuming that the channel is invariant over the duration of the
OFDM symbol and is in perfect synchronization, the received signal ycp is written as

ycp = xcp ⊗ h + ncp , (2)

where h = [h0 , h1 , · · · , h L−1 ] T is an L × 1 CIR vector generated by Gaussian random

variables, and ncp is a ( Nfft + Ncp ) × 1 vector, which is Additive White Gaussian Noise
(AWGN) with zero mean and variance σn2 .
When sampling for the received signal is completed, Nfft -point FFT is performed after
removing the CP for demodulation. Therefore, the FFT output for the received signal is
expressed as
Y = XH + w, (3)
Electronics 2023, 12, 4740 4 of 15

where X is an Nsc × Nsc diagonal matrix containing the transmitted signal. H is an Nsc × 1
CFR vector and w is an Nsc × 1 i.i.d. complex Gaussian random vector with zero mean and
variance σn2 . The n-th component of CFR Hn is expressed as [23]

L −1
Hn = FFT {h} = ∑ hl e− j2πln/Nfft , (4)
l =0
0 ≤ n ≤ Nfft − 1.

Since the transmission signal is concentrated on the pilot symbol-based method

through the reference signal, it is written with the received signal and CFR for the subcarrier
containing the pilot as
Z = SG + w̄, (5)
where Z and G are N × 1 vectors that are CFR and received signals for pilot subcarriers,
respectively. S is the N × N diagonal matrix of the reference signal according to Z, and w̄
represents N × 1 i.i.d. complex Gaussian random vectors with zero mean and variance σn2 .
Consequently, the channel estimation method is performed through Equation (5).

3. Preliminaries on Channel Estimation

3.1. LS/MMSE Method
Using the N × 1 reference signal vector S for the received signal Z, we perform the LS
method and estimate the channel as follows

ĜLS = S−1 Z
= S−1 (SG + w̄)
, (6)
= G + S−1 w̄
= G + w̃

where w̃ represents N × 1 i.i.d. complex Gaussian random vectors with vectors zero mean
and variance σn2 /E{SS† }. As shown in Equation (6), the LS method is one of the most
common approaches because of the simple calculation. However, it is difficult to make
precise channel estimation because w̃ remains in the estimated channel, and it does not use
the statistical information of the channel. Therefore, to overcome the limitations of the LS
method and minimize mean square errors, the MMSE method is performed as follows

ĜMMSE = RGZ R− 1
ZZ Z, (7)

where RGZ = E{GZ† } is an N × N cross-correlation matrix between the channel and

received signal, and RZZ = E{ZZ† } is the N × N auto-correlation matrix for the received
signal. The MMSE method may improve channel estimation accuracy by using the statistical
information of the channel. However, it has a higher computational complexity than the LS
method and requires statistical information of the channel. Additionally, there is a limitation
in using the MMSE method because it may be difficult to obtain channel information, and
accurate information is not guaranteed.

3.2. Existing ML Method

3.2.1. Structure
To compensate for the shortcomings of existing channel estimators, ML-based channel
estimation that performs as well as the MMSE method without utilizing the statistical
information of the channels is being studied. Figure 3 shows the simplest form of the ML
method [7] among the existing ML methods used to compare with the proposed method in
this paper. The ML-aided channel estimation is designed to minimize the MSE between
the actual channel and the estimated channel obtained by the LS method to overcome the
LS and MMSE methods, and the number of nodes in the input layer and output layer is
Electronics 2023, 12, 4740 5 of 15

2N. The number of hidden layers is 3, and the number of nodes in the hidden layers is
designed to be 4N. In detail, the input data use the estimated CFR ĜLS by the LS method,
and when these input data are input to the ML, the ML outputs a new estimated CFR ĜML
corresponding to ĜLS . Since the ML does not accept complex numbers, ĜLS must be input
by dividing it into real and imaginary numbers and ĜLS is also output divided into real
and imaginary numbers as follows

Re( ĜLS,d (0))

 
 Im( ĜLS,d (0)) 
..
 
Xin,d = ,
 
 . 
 Re( Ĝ
LS,d ( N − 1)) 
Im( ĜLS,d ( N − 1))
Re( ĜML,d (0)) (8)
 
 Im( ĜML,d (0)) 
..
 
Yout,d = ,
 
 . 
 Re( Ĝ
ML,d ( N − 1)) 
Im( ĜML,d ( N − 1))
d = {1, . . . , D },
where N is the length of the pilot signal, and Xin,d and Yout,d are the input and out-
put for the d-th data, respectively. Since the Yout,d output by the ML is divided into
real and imaginary numbers, it is converted to a complex number to obtain ĜML,d =
[ ĜML,d (0), . . . , ĜML,d ( N − 1)]T .

Figure 3. The structure of the existing ML-based channel estimator.

3.2.2. Activation Function

The m0 -th node of the k-th hidden layer performs the following calculations:

M
ok,m0 = f m0 (z) = f m0 ( ∑ wm xm + bm ),
m =1
(9)
k = {1, 2, 3},
m0 = {1, . . . , M0 },

where M is the number of inputs (m = 1, . . . , M ) to the m-th node in the previous layer,
and M0 is the number of nodes in the current hidden layer. wm is the m-th weight and bm is
the bias. f (·) is the activation function used to characterize the nonlinearity of the channel
Electronics 2023, 12, 4740 6 of 15

data, and ok,m0 is the output for the m0 -th node of the k-th hidden layer. In this model, the
activation function of the hidden layer adopts the tanh function as

ez − e−z
f (z) = , (10)
ez + e−z

where e is Euler’s number. The output ĜML,d for the d-th data is obtained in ML from the
input ĜLS,d and uses the loss function to minimize the MSE between the prediction channel
and the actual channel as follows
D N −1
1
L(W , B) =
DN ∑ ∑ k Gd (n) − ĜML,d (n)k2 , (11)
d =1 n =0

where Gd (n) is the actual channel value associated with ĜML,d (n), and W and B include
all the weights and biases, respectively. From a set of initial values, the weights and biases
are updated by minimizing the loss function shown in Equation (10).
Since ML-aided channel estimation is based on LS estimates, it minimizes the MSE
between actual channels, so it performs better than the LS method, but there are three
hidden layers, and each node is designed twice the input size, resulting in high complexity.
Therefore, we propose a low-complexity ML-based channel estimation where ML designed
the DNN with a single hidden layer. We also propose weight quantization that can save
memory stored by reducing the capacity of weights.

4. Proposed ML Method
4.1. Network Architecture
4.1.1. Structure
To compensate for the shortcomings of LS, MMSE, and existing ML-based channel
estimation, we propose a CIR estimation that is closer to the actual channel than the LS
method and requires no prior information, unlike the MMSE method. Figure 4 shows the
structure of the proposed ML-based channel estimation, which includes a DNN architecture.
To propose the low-complexity ML, it is designed with only one hidden layer and estimates
a CIR shorter than CFR.

Figure 4. The structure of the proposed ML-based channel estimator.

The receiver receives transmission signals coming through channels in various direc-
tions. Therefore, the reception signal can be expressed as the convolution product of the
channel impulse response and the transmission signal, and the first part of the reception
signal is affected by the reception signal of the previous symbol, resulting in Inter-Symbol
Interference. Accordingly, the communication system prevents Inter-Symbol Interference
by adding a CP in front of the transmission signal. In general, since the length of the
channel impulse response is smaller than the CP, ISI prevention is possible. Long pilot
signals are transmitted for more accurate channel estimation, but too-long pilot signals
generate overhead. The pilot signal has an appropriate length and thus occurs longer than
Electronics 2023, 12, 4740 7 of 15

the length of the channel impulse response. Consequently, the L-length CIR output from
the proposed ML is shorter than the N-length CFR output from the existing ML, and a
simpler ML can be designed.
It is necessary to distinguish between real and imaginary parts of the complex signal
because the current ML platform only allows real numbers. Therefore, the input data of the
ML for the d-th data Xin,d are given as follows

Re( Zd (0))
 
 Im( Zd (0))

..
 
Xin,d = . (12)
 
 . 
 Re( Zd ( N − 1)) 
Im( Zd ( N − 1))

The number of nodes in the output layer is 2L according to the length of CIR L, and
the output value of the ML for the d-th data Yout,d is given as follows
 
Re(ĥd (0))

 Im(ĥd (0))


 .. 

 . 

 Re(ĥd (l )) 
 
Yout,d = , (13)
 Im(ĥd (l )) 
..
 
 

 . 

 Re(ĥd ( L − 1)) 
Im(ĥd ( L − 1))

where ĥd (l )(l = 0, . . . , L) is the output of the neural network for the l-th tap of the d-th
data. The number of nodes in the hidden layer is less than the number of pilot signals N
and the largest is the power of 2, and it is represented by Q = 2blog2 ( N )c . Since the objective
of ML-based estimation is to minimize the MSE between the estimated channel and the
actual channel, the loss function L used in the training step is defined as follows

D L −1
1
L(W , B) =
DL ∑ ∑ khd (l ) − ĥd (l )k2 , (14)
d =1 l =0

where D is the number of data used for training, and hd (l ) is the actual channel value associ-
ated with ĥd (l ). From a set of initial values, the weights and biases are updated by minimizing
the loss function shown in Equation (13) with the forward and backward propagation.
The proposed method estimates the CIR, so the CIR is converted into the CFR to derive
the pilot symbol-based CFR ĜML as follows

ĤML = F Nfft · [ĥML , 0 Nfft − L ] T ,

(15)
ĜML = ĤML (idx(S)),

where 0 Nfft − L is the ( Nfft − L) × 1 zero vector, and F Nfft is the Nfft -point FFT matrix. idx(S)
refers to the location where the pilot signal is placed on the subcarrier.

4.1.2. Activation Function

Since there is only one hidden layer of the proposed method, nodes are defined
as follows
M
o m 0 = f m 0 ( z ) = f m 0 ( ∑ w m x m + bm ) . (16)
m =1

Since the purpose of the proposed method is to reduce complexity using the simple
DNN, the activation function of the DNN f (·) uses the tanh function shown in Equation (9)
Electronics 2023, 12, 4740 8 of 15

the closer the input is to zero. The greater the differential value, the more easily it is able to
converge and complete the training quickly. The tanh function is shown in Figure 5. As
shown in Figure 5a, the output of the function is in the [−1, 1] interval and the median is 0,
so there is no bias. Additionally, since the gradient is mostly steep and can be both positive
and negative, it quickly converges to the optimal value and has almost no gradient loss, as
shown in Figure 5b.

(a) (b)

Figure 5. (a) The tanh function, (b) derivative of tanh function.

4.1.3. Complexity Analysis

In this paper, we compare computational complexity through the number of multipli-
cations for the existing methods and the proposed method. The computational complexity
of the existing ML and the proposed ML add the LS method for the input data ĜLS and
the Nfft point FFT to transform from CIR to CFR, respectively. The results are shown in
Table 2, where Q = 2blog2 ( N )c is the largest power of 2, while it is less than N. Here, Nfft
is the FFT size. In addition, N and L are the pilot signal size and the length of the CIR,
respectively. These results show that the existing ML is lower than the MMSE method and
slightly higher than the LS method, but the proposed ML is lower than the LS method as
well as the MMSE method.

Table 2. Computational complexity of ML-based estimators with LS and MMSE.

The Number of Computational

Algorithm
Multiplications/Inversions Complexity
LS N2 + N O( N 2 )
MMSE N 3 + 3N 2 O( N 3 )
N 2 + N + 2N × 4N + 3 × (4N × 4N ) + 4N × 2N
Existing ML O( N 2 )
= 65N 2 + N
2N × Q + Q × 2L + N f f t × L
Proposed ML O( QN )
= 2Q( N + L) + N f f t × L

4.2. Quantization Method

The trained ML performs a quantization of 32-bit floating-point weights w to 8-bit
integer-point weights for the interval [−128, 127], as shown in Figure 6. For quantization,
values that are 32 bits are divided into a certain range to quantize them into 8 bits. In
this figure, different proximity weight values, such as wk−1 and wk , are quantized to
the same value if they belong to the same interval when quantized to 8 bits. Different
weight values within the interval are quantized to the same value, so performing inverse
quantization to use ML converts them to the same weight value. Therefore, when the
weights go through quantization, they are set to a value different from the original value,
resulting in performance loss. The maximum value wmax and minimum value wmin of
Electronics 2023, 12, 4740 9 of 15

w are stored in memory for dequantization. Then, when using the proposed ML for
channel estimation, quantized weights must be applied to ML after dequantization. The
dequantization normalizes the interval [−128, 127] with stored wmax and wmin to derive
the interval [wmin , wmax ].

Figure 6. The structure of the weight quantization algorithm.

In the case of weight quantization for memory saving, wmax and wmin must be stored
because it is to be restored based on them. However, if the length of the CIR is L = 1
and L = 6, then each cumulative distribution function (CDF) of the proposed ML for
the weights w is shown in Figure 7. As shown in this figure, the probability that the
value of |w| exists below 0.5 is F (w) = P(|w| ≤ 0.5) > 0.99. Therefore, almost all the
weights are assigned within the interval [−0.5, 0.5], and the dequantization step uses the
interval [−0.5, 0.5] to dequantize the interval [−128, 127] instead of using wmax and wmin
for memory saving.

(a) L = 1 (b) L = 6

Figure 7. CDF of weights for the proposed ML.

5. Simulation Analysis
5.1. Simulation Environment
In this paper, the environment for channel estimation is based on the pilot signal
generated through the 5G NR Sounding Reference Signal (SRS) [24]. In 5G NR, the OFDM
system is almost the same as the OFDM system in LTE [25], so it is possible to follow the
system model in Section 2. SRS is generated based on the Low Peak to Average Power
Ratio (PAPR) sequence as follows

(α,δ)
ru,v (n) = e jαn r̄u,v (n), RS
0 ≤ n ≤ Msc , (17)
Electronics 2023, 12, 4740 10 of 15

where Msc RS = mN RB /2δ is the length of the SRS, and m is the number of resource blocks
sc
(RBs) occupied by the SRS. Nsc RB is the number of subcarriers contained in one RB, and

δ is δ = log2 (KTC ) according to transmission comb number KTC ∈ {2, 4, 8}. The base
sequence r̄u,v (n) depends on the group number u and the number of base sequences v, and
multiple orthonormal SRS can be generated using different cyclic shift α even with the
same base sequence.
For MZC ≥ 3Nsc RB , the base sequence r̄ ( n ) (0 ≤ n ≤ M RS ) is given by
u,v sc

r̄u,v (n) = xq (n modNZC ),

πqm(m+1) (18)
−j NZC
xq (m) = e ,

where q is given by
q = bq̄ + 1/2c + v · (−1)b2q̄c ,
(19)
q̄ = NZC · (u + 1)/31.

The length NZC is given by the largest prime number such that NZC < Msc RS .

To evaluate the performance of the proposed ML-based channel estimation in this

paper, an algorithm for channel estimation was designed using MATLAB (R2021b)-based
simulation [26], and the parameters used in the simulation are shown in Table 3. The
correlation matrices of the MMSE method design an ideal MMSE using the actual CFR G.
Parameters required for the proposed ML are shown in Table 4. As an additional technique,
ML utilizes an Adam optimizer to quickly reach global optimum points without converging
to local optimum points [27]. The structure of the layer in the proposed ML is designed as
shown in Table 5. The performance of the ML is verified after the training is completed. In
this simulation, the CIR was set in consideration of the AWGN channel with L = 1 and
the multipath channel with L = 6, wherein the power of each component of the CIR is the
same and the sum is 1. The trained ML performs weight quantization for the weights in
the ML, and quantized weights are stored in memory.

Table 3. The parameters for OFDM system.

Parameters Values
SRS size ( N ) 48
Subcarrier size ( Nsc ) 216
FFT size ( Nfft ) 256
Tap size ( L) 1, 6
Channel model Gaussian channel
Noise model Gaussian noise
SNR [−10 : 5 : 20]

Table 4. The parameters for proposed ML.

Parameters Values
Number of hidden layer 1
Input layer size (2N ) 96
Hidden layer size ( Q) 32
Output layer size (2L) 2, 12
Batch size 8
Learning rate 10−4
Training epochs 100
Activation function tanh
Optimizer Adam
Loss function Mean squared error
Electronics 2023, 12, 4740 11 of 15

Table 5. The structures of each layer for the proposed ML.

Layers 1Tap DNN 6Tap DNN

Nodes f (·) Nodes f (·)
Input layer 96 - 96 -
Hidden layer 32 tanh 32 tanh
Ouput layer 2 - 12 -

The performance metric between the actual channel and the estimated channel was ver-
ified using the MSE defined as follows in Equation (20) for each signal-to-noise ratio (SNR)
I N −1
1
MSE =
IN ∑∑ kGi (n) − Ĝi (n)k2 , (20)
i =1 n =0

where Gi (n) is the actual channel for the n-th component of the i-th iteration, and Ĝi (n) is
also the estimated channel for the component corresponding to Gi (n).

5.2. Simulation Results

5.2.1. Comparison between Existing Methods and Proposed Method
Figure 8 compares the MSE performance of the proposed channel estimation method
with the two lengths of the CIR scenarios. Figure 8a,b show the length of the CIR at L = 1
and L = 6, respectively. The performance of the proposed method is compared to the
LS, MMSE, and existing ML methods, and all methods improve their performance as the
SNR increases. In both scenarios, the LS methods do not take advantage of the statistical
information of the channel, resulting in the worst MSE performance. On the other hand,
the performance of the MMSE method is better than the LS method in both scenarios,
because the MMSE method uses the statistical information of the channel. The existing
ML methods and the proposed methods performed better than the LS methods in both
scenarios. However, the performance of the existing ML method is better than the MMSE
method below about 7 dB in the L = 1 scenario, but it reverses above 7 dB. And it is worse
than the MMSE method at all the SNRs in the L = 6 scenario. The performance of the
proposed method without performed weight quantization is better than the MMSE method
at all the SNRs in the L = 1 scenario but is worse than the MMSE method at all the SNRs in
the L = 6 scenario. Both the existing and proposed MLs suffer performance degradation
depending on the length of the CIR, which is estimated to reduce ML’s computational
accuracy as the received signal becomes complex due to the convolutional multiplication
of the transmitted signal. Furthermore, the performance degradation is expected to be due
to the fact that the received signal information does not increase proportionally with the
length of the CIR, but rather the longer CIR must be inferred from the same amount of
received signal information, leading to more challenging results.

(a) L = 1 (b) L = 6

Figure 8. Performance comparison between existing methods and the proposed method.
Electronics 2023, 12, 4740 12 of 15

Meanwhile, it outperforms the existing ML method in both scenarios like the proposed
method without the quantization. The quantized ML methods have no differences in per-
formance from the proposed ML methods without performing weight quantization below
the SNR of 10 dB in both scenarios. It is also observed that there are slight performance
differences in the SNR regime which is greater than or equal to 10 dB, but it remained better
than the existing ML methods.

5.2.2. The Number of Hidden Layers

To ensure the reliability of the proposed method of designing only one hidden layer,
the proposed method is designed with three hidden layers like the existing ML method
and compared with the one-hidden-layer method. As shown in Figure 9, there are little
differences between proposed algorithms with 1 and 3 hidden layers in both scenarios.
Therefore, it can be seen that the proposed method can be efficiently designed to complete
the training quickly by reducing the complexity of ML by setting one hidden layer rather
than slightly increasing the performance by setting a lot of hidden layers. Although the
performance loss for weight quantization appears to occur similarly in both methods, the
performance becomes unstable as the number of hidden layers increases, as shown in the
L = 6 scenario.

(a) L = 1 (b) L = 6

Figure 9. Performance comparison between proposed algorithms with 1 and 3 hidden layers.

5.2.3. ML Robustness to Other SNRs

As shown in Figure 10, the model of the proposed method was trained with only
the set SNR, and the performance is compared by entering data for each SNR. In both
scenarios, there is little performance loss compared to the correct trained models for each
SNR. Moreover, in the L = 1 scenario, below about 10 dB, the model trained with 10 dB
data performs better than that trained with 20 dB data, but above about 10 dB, the model
trained with 20 dB data performs better than that trained with 10 dB data. Consequently,
the proposed method is robust against other SNR data, as there is little performance loss
even if SNR data different from that of the trained data are inputted.

(a) L = 1 (b) L = 6

Figure 10. Performance based on different SNR data than the trained SNR data.
Electronics 2023, 12, 4740 13 of 15

6. Conclusions
In this paper, we provided the light-weighted ML approach to channel estimation for
NR systems. The main intuition was that there exists an equivalence between the CIR in the
time domain and its corresponding CFR in the frequency domain. Based on the CFR rather
than the CIR, it was shown that the more light-weighted ML model can be established
in comparison with the existing ML-based channel estimator. Specifically, regardless of
the number of CIR’s channel taps, it was observed that the ML model configuration with
one input layer, one hidden layer, and one output layer was sufficient to learn the channel
estimation, which was confirmed from the computational complexity comparison in Table 2.
Furthermore, the proposed light-weighted ML-based channel estimator was shown to be
robust when the quantized weights were applied to reduce memory overhead, which
opened the availability for practical use. From the perspective of the MSE performance,
the proposed ML-based estimator has better performance than the existing ML-based
channel estimator. One interesting thing is that the proposed ML-based estimator had
better performance even than MMSE when the number of channel taps is equal to 1,
which showed the effectiveness of the proposed one especially for poor scattering channel
environments. Instead of naive use of the proposed ML-based estimator training for each
SNR, it was observed that with a trained proposed ML-based estimator, the specific SNR
is sufficient to cover all the SNR regimes without significant performance degradation.
As a further work, the proposed ML-based estimator can be naturally extended to CFR
interpolation based on the inference of the CIR, which can be further elaborated.

Author Contributions: Conceptualization, H.W.L. and S.W.C.; methodology, H.W.L. and S.W.C.;
software, H.W.L.; validation, H.W.L. and S.W.C.; formal analysis, H.W.L. and S.W.C.; investiga-
tion, H.W.L.; resources, H.W.L.; data curation, H.W.L.; writing—original draft preparation, H.W.L.;
writing—review and editing, S.W.C.; visualization, H.W.L.; supervision, S.W.C.; project adminis-
tration, H.W.L. and S.W.C.; funding acquisition, S.W.C. All authors have read and agreed to the
published version of the manuscript.
Funding: This work was supported by the Institute of Information & Communications Technology
Planning & Evaluation (IITP) grant funded by the Korea government (MSIT). (No. 2021-0-00165,
Development of 5G+ Intelligent Base Station Software Modem).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data are contained within the article.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:
3GPP 3rd Generation Partnership Project
5G 5th Generation
AWGN Additive White Gaussian Noise
CFR Channel Frequency Response
CIR Channel Impulse Response
CNN Convolutional Neural Network
CP Cylic-Prefix
DMRS DeModulation Reference Signal
DNN Deep NN
FFT Fast Fourier Transform
IFFT Inverse FFT
i.i.d. Independent and Identically Distributed
LSTM Long Short-Term Memory
ML Machine Learning
MNN Multi-Layer NN
Electronics 2023, 12, 4740 14 of 15

MSE Mean Square Error

NR New-Radio
OFDMA Orthogonal Frequency Division Multiple Access
PAPR Peak-to-Average Power Ratio
SRS Sounding Reference Signal
ZP Zero Padding

References
1. Hou, X.; Zhang, Z.; Kayama, H. DMRS Design and Channel Estimation for LTE-Advanced MIMO Uplink. In Proceedings of the
2009 IEEE 70th Vehicular Technology Conference Fall, Anchorage, AK, USA, 20–23 September 2009; pp. 1–5.
2. Wang, Y.; Zheng, A.; Zhang J.; Yang, D. A novel channel estimation algorithm for sounding reference signal in LTE uplink
transmission. In Proceedings of the 2009 IEEE International Conference on Communications Technology and Applications,
Beijing, China, 16–18 October 2009; pp. 412–415.
3. Bertrand, P. Channel Gain Estimation from Sounding Reference Signal in LTE. In Proceedings of the 2011 IEEE 73rd Vehicular
Technology Conference (VTC Spring), Budapest, Hungary, 15–18 May 2011; pp. 1–5.
4. Xia, X.; Zhao, H.; Zhang, C. Improved SRS design and channel estimation for LTE-advanced uplink. In Proceedings of the 2013
5th IEEE International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications,
Chengdu, China, 29–31 October 2013; pp. 84–90.
5. Tran, H.; Mai, T.-A.; Dang, S.; Ngo, H.-A. Large-scale MU-MIMO uplink channel estimation using sounding reference signal.
In Proceedings of the 2018 2nd International Conference on Recent Advances in Signal Processing, Telecommunications &
Computing (SigTelCom), Ho Chi Minh City, Vietnam, 29–31 January 2018; pp. 107–110.
6. Balevi, E.; Doshi, A.; Andrews, J.G. Massive MIMO Channel Estimation With an Untrained Deep Neural Network. IEEE Trans.
Wireless Commun. 2020, 19, 2079–2090. [CrossRef]
7. Le Ha, A.; Van Chien, T.; Nguyen, T.H.; Choi, W. Deep Learning-Aided 5G Channel Estimation. In Proceedings of the 2021 15th
International Conference on Ubiquitous Information Management and Communication (IMCOM), Seoul, Republic of Korea, 4–6
January 2021; pp. 1–7.
8. Mei, K.; Liu, J.; Zhang, X.; Cao, K.; Rajatheva, N.; Wei, J. A Low Complexity Learning-Based Channel Estimation for OFDM
Systems With Online Training. IEEE Trans. Commun. 2021, 69, 6722–6733. [CrossRef]
9. Ma, X.; Gao, Z. Data-Driven Deep Learning to Design Pilot and Channel Estimator for Massive MIMO. IEEE Trans. Veh. Technol.
2020, 69, 5677–5682. [CrossRef]
10. Jiang, P.; Wen, C.-K.; Jin, S.; Li, G. Y. Dual CNN-Based Channel Estimation for MIMO-OFDM Systems. IEEE Trans. Commun. 2021,
69, 5859–5872. [CrossRef]
11. Le, H.A.; Van Chien, T.; Nguyen, T.H.; Choo H.; Nguyen, V.D. Machine learning-based 5G-and-beyond channel estimation for
MIMO-OFDM communication systems. Sensors 2021, 21, 4861. [CrossRef] [PubMed]
12. Dayi, A.B. Improving 5G NR Uplink Channel Estimation with Artificial Neural Networks: A Practical Study on NR PUSCH
Receiver. In Proceedings of the 2022 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom),
Sofia, Bulgaria, 6–9 June 2022; pp. 129–134.
13. Liao, Y.; Hua Y.; Cai Y. Deep Learning Based Channel Estimation Algorithm for Fast Time-Varying MIMO-OFDM Systems. IEEE
Commun. Lett. 2020, 24, 572–576. [CrossRef]
14. Bai, Q.; Wang, J.; Zhang, Y.; Song, J. Deep Learning-Based Channel Estimation Algorithm Over Time Selective Fading Channels.
IEEE Trans. Cognit. Commun. Netw. 2019, 6, 125–134. [CrossRef]
15. Yang, Y.; Gao, F.; Ma, X.; Zhang S. Deep Learning-Based Channel Estimation for Doubly Selective Fading Channels. IEEE Access
2019, 7, 36579–36589. [CrossRef]
16. Jebur, B.A.; Alkassar, S.H.; Abdullah, M.A.M.; Tsimenidis C.C. Efficient Machine Learning-Enhanced Channel Estimation for
OFDM Systems. IEEE Access 2021, 9, 100839–100850. [CrossRef]
17. Han, S.; Mao, H.; Dally, W.J. Deep compression: Compressing deep neural networks with pruning, trained quantization and
Huffman coding. arXiv 2015, arXiv:1510.00149.
18. Gholami, A.; Kim, S.; Dong, Z.; Yao, Z.; Mahoney, M.W.; Keutzer, K. A Survey of Quantization Methods for Efficient Neural
Network Inference. arXiv 2021, arXiv:2103.13630.
19. van de Beek, J.-J.; Edfors, O.; Sandell, M.; Wilson S.K.; Borjesson, P.O. On channel estimation in OFDM systems. In Proceedings of
the 1995 IEEE 45th Vehicular Technology Conference. Countdown to the Wireless Twenty-First Century, Chicago, IL, USA, 25–28
July 1995; pp. 815–819.
20. Hsieh, M.-H.; Wei, C.-H. Channel estimation for OFDM systems based on comb-type pilot arrangement in frequency selective
fading channels. IEEE Trans. Consum. Electron. 1998, 44, 217–225. [CrossRef]
21. Morelli, M.; Mengali, U. A comparison of pilot-aided channel estimation methods for OFDM systems. IEEE Trans. Signal Process.
2001, 49, 3065–3073. [CrossRef]
22. Coleri, S.; Ergen, M.; Puri, A.; Bahai, A. Channel estimation techniques based on pilot arrangement in OFDM systems. IEEE Trans.
Broadcast. 2002, 48, 223–229. [CrossRef]
Electronics 2023, 12, 4740 15 of 15

23. Hajizadeh, R.; Mohamedpor, K.; Tarihi, M.R. Channel Estimation in OFDM System Based on the Linear Interpolation, FFT and
Decision Feedback. In Proceedings of the 18th Telecommunication forum TELFOR 2010, Serbia, Belgrade, 23–25 November 2010.
24. 3GPP TS 38.211 v17.3.0; NR Physical Channels and Modulation. European Telecommunications Standards Institute: Sophia
Antipolis, France, 2022.
25. Kumar T.A.; Anjaneyulu, L. Channel Estimation Techniques for Multicarrier OFDM 5G Wireless Communication Systems.
In Proceedings of the 2020 IEEE 10th International Conference on System Engineering and Technology (ICSET), Shah Alam,
Malaysia, 9 November 2020; pp. 98–101.
26. Cho, Y.S.; Kim, J.; Yang, W.Y.; Kang, C.G. MIMO-OFDM Wireless Communications with MATLAB; John Wiley & Sons: Hoboken, NJ,
USA, 2010.
27. Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv 2014, arXiv:1412.6980.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.