Highly Accurate Technique for CO-OFDM Channel Estimation

Technique Using Extreme Learning Machine (ELM)

NISHA MARY JOSEPH, PUTTAMADAPPA C.

Department of Electronics and Communication Engineering, Dayananda Sagar University,

Kudlu gate, Hosur, Bangalore,

INDIA

Abstract: - In wireless systems, channel estimation is considered a problematic technology, due to the fact of

and receiver sides and the relatively long OFDM symbol duration compared to that of single carrier

communications. An Extreme Learning Machine (ELM) with Pilot Assisted Equalization (PEM) is proposed

for compensation of impairments caused by fibre nonlinearity in coherent optical communication systems.

Channel estimation using ELM and the value of distortion is sent to the OSTBC receiving end based on the

distortion information the data is decoded and pilot data is removed. FFT is applied to the data and QPSK

demodulation is done in the data to get its original form. In addition, the article utilized a free-space optical

communication system of multi-input multi-output orthogonal frequency division multiplexing (MIMO-

OFDM) with a modified receiver structure. Simulation reveals that the proposed model exhibits significant

BER (0.0112) performance and provides better spectral efficiency as compared with conventional systems and

less computational complexity. This suggested that the proposed method shows better performance by using the

Received: April 4, 2022. Revised: January 6, 2023. Accepted: February 12, 2023. Published: March 9, 2023.

1 Introduction

Over the past few decades, global network traffic

has grown explosively due to the demands for

higher bandwidth and faster connections of various

multimedia and data services (e.g. big data, cloud

computing, streaming video, Internet of Things,

machine-to-machine communication, and remote

transmission medium in fiber-to-the-home (FTTH)

networks, metropolitan area networks (MANs),

backbone networks, and transoceanic

communications, [2]. Due to the wide application of

fibre-optic networks, optical fibres exist anywhere

in modern society, and fibre can be exploited for

more functions than data transmission. In addition to

acting as the transmission medium in fibre-optic

networks, optical fibre can also act as the sensing

medium in optical fibre sensors, [3]. The fibre

(OFDM) systems with complex and long-haul

transmission, [4]. Orthogonal frequency-division

multiplexing (OFDM) is a widely used

modulation/multiplexing technology in wireless and

data communications. With recent advances in high-

speed CMOS technologies and optical modulation

and detection technologies, optical OFDM at 40-

Gb/s or even 100-Gb/s information rate becomes

feasible. Together with digital coherent detection,

receiver sensitivity as coherent single-carrier

transmission, [5]. A key feature of CO-OFDM is its

transmitter to facilitate channel estimation, which

provides crucial information about the transmission

channel and enables efficient digital compensation

of optical transmission impairments such as

chromatic dispersion (CD) and polarization-mode

dispersion (PMD). High peak-to-average power

ratio (PAPR) remains the main challenge of

(OFDM) technology and multi-input multi-output

(MIMO) technology are widely used in wireless

communication systems, improving spectral

efficiency, data transmission quality, and system

capacity, [7]. In MIMO-OFDM wireless

communication systems, the channel state

information (CSI) obtained by channel estimation

techniques is critical, as it has a significant impact

on the accurate implementation of coherent

detection and decoding and directly guides the

sub-1 ms round-trip latency), [9]. Especially in

channel estimation, an increasing number of channel

coefficients and long pilot sequences need to be

trained, and increasingly stringent latency

constraints are imposed. These pose growing

challenges to existing channel estimation based on

electronic processors in terms of computing power

proportional to parallelism, data storage, and data

transmission, [10]. Specifically, channel estimation

entails large-scale complex matrix computations,

(e.g., serial-to-parallel conversion and fast Fourier

transform), there is additional latency of

microseconds in signal access, processing, storage,

and transmission, [11]. Thus, key performances of

channel estimation still have room for improvement

latency of the entire channel estimation process and

improve the computational speed and parallelism of

channel estimation.

interference approximation method (IAM) have

been proven to be effective in suppressing the IMI

with low complexity, [12]. For the IAMs, increasing

the power of the pseudo pilot could decrease the

effects of amplified spontaneous emission (ASE)

noise and other frequency domain residual errors on

resulting in non-optimal channel estimation

accuracy. Recently, Xi Fang suggested the channel

estimation model based on the combined phase

offset (CMPO). For the CMPO method, the first and

the third pilot blocks are also loaded with pilots.

With this consideration, a pilot structure called E-

IAM-C was proposed for OFDM/OQAM,

improving the pseudo pilot power compared with

the IAMs and PHO, [13]. However, for the E-IAM-

C method, matrix inversion is required due to the

high receiver sensitivity and spectral efficiency, and

robustness against dispersion. However, the impact

of the optical channel is a major limiting factor in

CO-OFDM systems when the speed rate is up to

100 Gbit/s or higher. Therefore, it requires to use of

limited attention in the field of CO-OFDM systems

in recent years. Many channel estimation methods

were reported in the CO-OFDM system, such as

least square (LS), maximum likelihood (ML), linear

minimum mean square error (LMMSE) algorithms,

and deep neural network, [15]. Recently, a neural

network, [5] was used as a nonlinear model to

approximate the relationship between the channel

impulse response and its corresponding subcarrier

index. Most semi-blind estimation algorithms

trained by using gradient-based learning algorithms

extensively, so their learning speed is in general far

slower than required and these learning algorithms

exist local minimum problems.

technology which overcomes some challenges faced

by other techniques has recently attracted more and

more attention. Consequently, the article presented

an improved channel estimation based on the ELM

algorithm with CO-OFDM with MIMO optical

system is its sensitivity to fibre nonlinearity. Pilot

Assisted Equalization model with ELM has been

demonstrated capable of compensating fiber

nonlinear distortion in a QPSK optical

communication system. Increasingly extensive

application scenarios and exponentially growing

data volumes of MIMO-OFDM systems have

imposed greater challenges on the speed, latency,

and parallelism of channel estimation based on

electronic processors. However, in optical

response and predict the actual channel response

from the receiver.

section 2, the literature survey of the study is

defined. The problem definition and motivation of

improvement in transmission distance by 3 km with

a previously used WDM-based multi-beam FSO

system. It performs better in all respects with BER

distance of 7 km under atmospheric turbulences at

35 dB attenuation. Further, a significant reduction in

geometric loss has been achieved for a given

distance in varying turbulence fluctuations and the

numeric values have been substantiated with Rytov

variance.

systems. Compared with the full-loaded frequency-

domain channel estimation (FL-FD) method, the

CISFDA method promotes the system robustness

against the ASE noise and fibre nonlinear effect

significantly. The proposed method is validated by

numerical Monte Carlo simulations of the PDM

CO-OFDM/OQAM system. Our new algorithm

outperforms the FL-FD method in the 36 Gb/s PDM

OFDM CO-OFDM/OQAM system, and the

transmission distance is improved by 500 km with

transfer function (TF) in the frequency domain by

LS to effectively reduce IMI. Through a series of

simulation test data, it can be found that this

algorithm can effectively improve the overall

performance of the system. Compared with the

improved by employing matched filter before DSP

which has the property of noise rejection i.e.

waveforms can be detected in presence of a jammer.

Further, a comparison of the proposed system is

performed with QPSK-IsOWC, DP-QPSK-IsOWC,

and DP-QPSK-CO-OFDM-IsOWC systems in terms

of log BER, error vector magnitude, received power,

and optical signal to noise ratio. It is observed that

the proposed system can cover 15,900 km at 100

Gbps at a targeted log BER of − 2.42 and provides

has not been reported in the past.

scheme to detect and locate physical-layer

eavesdropping. To improve the efficiency and

accuracy of eavesdropping detection, both Optical

Performance Monitoring (OPM) data and eye

diagrams are adopted as input data. Three different

ML classifiers are designed and tested to realize

eavesdropping detection, location, and split ratio

recognition, respectively. To demonstrate the

addition, different splitting ratios are considered,

including 95/5, and 90/10. Results indicate that the

ML scheme achieves 100% and 92.76% accuracy in

eavesdropping detection and location, respectively.

detection (IM/DD) optical Fast-OFDM systems

were examined by [22], and they developed

combined channel estimation and digital

linearization schemes. To transmit the 2PAM

signals, the odd SCs of the training sequences were

increase the receiver sensitivity in comparison with

the conventional IM/DD Fast-OFDM. A least mean

square (LMS) algorithm has been proposed by [23],

to evaluate the phase noise. The advantage of this

algorithm is, the bit error rate is improved and the

filter (AEKF) blind scheme has been developed

based on the theory of the Kalman filter and the PN

model to achieve flexibility in dynamic networks. A

change in the identification of blind phase with a

feedback loop has a temporal complexity that is just

1/3 that of a commercial laser with a linewidth of

200 kHz.

analysis based on the weight function (FICA-WF)

for blind interference cancellation. The existing

models such as parallel factor analysis, joint parallel

factor analysis, STBC-m-MIMO-OFDM, and

MMSE-CMA-DFCE obtained SNR ranging from 10

to 20 dB, whereas the proposed model obtained

better SNR of 9.02 dB for the FastICA-WF. [26]

proposed a novel blind CE method for vehicular

VLC to improve CE accuracy based on the

exploitation of the channel statistics derived, by

zenith angle, and ambient light. Then, this channel

characteristic is exploited in the blind CE to

improve its accuracy with a two-step estimation of

the normalization factor. Extensive simulations at

different vehicle speeds show that the proposed

method outperforms the pilot-based and

superimposed training-based CE methods in terms

of spectral efficiency both for all modulation

schemes and at all relative speeds. The proposed

blind channel estimation (CE) method provides a

Biased Optical Orthogonal Frequency Division

Multiplexing (DCO-OFDM). [27] examine the deep

neural network (DNN) layers created from long-

short-term memory (LSTM) for detecting the

signals by learning the received signal as well as

survey, the study utilized the Optical fibre

transmission using coherent OFDM with FSO

optical communication in the means of MIMO

networks using an Extreme Learning Machine

(ELM). Table 1 shows the comparative analysis of a

recent work based on optical fibre communication.

the subcarriers and hence the signals found in the

sub-channels may interfere with each other in the

OFDM system. The OFDM system's vulnerability to

frequency deviation is one of its fundamental flaws.

Some advantages of the optical IMDD-OQAM-

OFDM system over the intensity modulation and

direct detection (IMDD) method include high

spectrum efficiency due to a higher side lobe

suppression ratio provided by a specially designed

filter bank, the viability of asynchronous

and has real field orthogonality built in, the phase

noise is made trickier and more challenging to

considered an essential module, to enhance the

orthogonal frequency division multiplexing

(OFDM) system execution. The high spectral

efficiency can be acquired because there is no need

for the time-frequency domain well-localized pulse

shapes and CP. The frequency offset and robustness

(CPE) and the complex domain (C-ELM)

compensation in terms of the BER over an additive

white Gaussian noise channel and several laser

oscillators. Some disadvantages were categorized

based upon both techniques namely: An extra

quadrature amplitude modulation with 16-ary.

and optical spectral efficiency, dispersion

insensitivity, high optical signal-to-noise ratio

(OSNR) sensitivity, and computation efficiency.

Channel estimation and equalizations are the most

important process on the receiver side of all kinds of

communication networks to estimate the

characteristics of the channel. In this work, a new

technique is proposed to perform the channel

estimation technique to improve the BER

performance for high-speed data communication

optimal design. Figure 1 illustrates the block

diagram of the proposed work.

dispersion. Further, the research utilized coherent

multiple-input multiple-output (MIMO) architecture

for optical wireless communications (OWCs) to

mitigate atmospheric turbulence effects. The beam

intensity profile was measured for investigating the

temporal nature of atmospheric turbulence on the

optical beam. Transmitter optical signals operate at

distinct carrier frequencies to allow the received

optical signals to be separately processed. It

increases the channel capacity of the system almost

Coherent detection is usually considered necessary

coherent detection, this has the same sensitivity as

binary phase-shift keying (BPSK) but doubles the

spectral efficiency. However, it has been shown that

OFDM is sensitive to phase noise. Specifically, the

phase noise will cause common phase error (CPE)

noise and inter-carrier interference (ICI). The CPE

can be estimated by the pilot-assisted (PA)

approach. At the CO-OFDM receiver after

removing CP and implementing the FFT operation,

it is followed by a channel estimation operation.

performance with high spectral efficiency and less

computational complexity.

4.1 Coherent Optical OFDM Network

The optical domain OFDM coherent identification

and modulation are used to accomplish CO-OFDM.

It has good receiver sensitivity and spectral

efficiency performance. A coherent optical OFDM

system's block diagrams are shown in Figure 2. The

random sequence generator will generate the

random sequence in the proposed system. To

System

Further, QPSK modulation is performed on the

sequence that is generated by the random sequence

generator. By swapping the stage of a constant

frequency reference signal, the data can be carried

out in a digital modulation process. The modified

signal is employed by the IFFT as the signal from its

original domain is transferred by Fourier analysis

transforms, which is frequently time or space, to a

distortion and IQ mismatch factors. Data along with

the pilot is sent to the OSTBC encoder and the

encoded data is sent to the optical channel.

channel estimation using ELM and the value of

distortion

is sent to the OSTBC receiving end

based on the distortion information the data is

decoded and pilot data is removed. FFT is

implemented in the data and QPSK demodulation is

done in the data to get its original form.

data pipes, the serial data input is

subcarriers were inserted gradually. To establish the

digital time-domain signal, the inverse FFT is used,

which is then merged with a cyclic prefix (CP) to

reduce the interruption of the inter-symbol caused

by multipath channels. By employing an ideal

digital-to-analogue converter (DAC) with a correct

sampling rate, such as an OFDM transmitter the

and

by

. The subcarrier index

needs

to be modulated

or unmodulated

. As

indicated, reference tones

are gradually

between the nearby subcarriers which will be

orthogonal, in which the time taken by the one

OFDM symbol is denoted

there was a difference noted in the phase noise in

can be rewritten as

amplitude,

and

the noise present in the

laser phase.

is the noise of frequency with

heterodyning the optical signal in which the

which is still found in the receiver and used to detect

frequency synchronization are assumed, hence the

CP is rejected and FFT is executed. The

arrived information symbol can be stated in

equation (8).

represents the RF phase noise dc

distance

(9)

obtained from discrete time. All the subcarriers of

matrix form as:

as described in the introduction. High performance

and low complexity can be defined through this

equalization method. The average power of the

constellation derives from the strength of pilot

subcarriers and for the sake of simplicity, the pilot’s

quadrature component is set to 0. In the PAE

technique, in a single constellation point, all the

other reference tones are present. The phase noise

for OFDM symbols can be decreased by the

equalizer especially when ICI dominates the CPE.

antennas, the receiver contains

all the symbols that were transmitted share

subcarriers. Figure 3 shows the MIMO-based

communication in optical channels. The sequence of

frequency-domain transmitted from the

between

receive antenna for the

, the

denotes the

signal transmitted on the

The CSI matrix

represents the block diagonal axis.

approach. Furthermore, in a Trans impedance

amplifier (TA), a positive-intrinsic-negative (p.i.n.)

photodetector is set up for the receiver's

implementation. By the recurrence of MIMO, the

photodetector output scheme

in equation (13).

scintillation experienced by plane waves and for