DOI: https://doi.org/10.21203/rs.3.rs-2487057/v1
The high demand for wireless communication and limited spectral power causes the conventional orthogonal multiple access approach ineffective for 5G communications. Thus, to specify the spectral inefficiency Multiple-input-multiple-output and non-orthogonal multiple access (MIMO-NOMA) were introduced. Here, MIMO and NOMA are integrated to earn to improve the channel capacity and spectral efficiency. However, the high Bit Error Rate (BER) and computational complexity in NOMA_MIMO due to successive interference cancellation (SIC) reduces the system performance for edge user. Thus, different channel estimation techniques are developed in the past to overcome these issues. But still, they face challenges in complexity and error rate. Hence, a novel hybrid Whale optimization algorithm with a Radial Basis Function Neural Network-based channel estimation method (WOA-RBFNN) was proposed in this article. The developed model estimates the path for data transmission for edge user and tunes the channel parameters till it attains their optimal value. The optimal fitness function in the proposed model offers the finest system performances in terms of Bit Error rate (BER), throughput, etc. Furthermore, the results are estimated and compared with the existing techniques for validation purposes. The comparative analysis proves that the developed model earned better performances than the existing ones especially for edge users.
NOMA is the most promising antenna technology in fifth-generation (5G) communication, as it offers high network coverage [1]. The major challenge in orthogonal multiple access (OMA) is, it provides less spectral power when the resources are allocated to users under poor channel conditions [2]. In NOMA, the spectral problem is minimized by performing successive interference cancellation (SIC) [3], and superposition coding (SC) [4]. The utilization of SC and SIC enables the network users to share the resources with different channel conditions [5]. Hence, the data coverage, and network are high in NOMA. Moreover, an optimal balance between the user and the spectral efficiency is attained in NOMA through effective power allocation and user pairing techniques [6]. In addition, another popular approach to enhance the network and spectral efficiency in cellular networks is multi-input and multi-output (MIMO) [7]. In MIMO, the spectral power is improved by utilizing data in the spatial dimension [8]. Furthermore, a multiuser MIMO downlink channel is achieved in communication networks by dirty-paper coding (DPC) [9]. Still, the execution of DPC is very complex and consumes more time. Moreover, DPC itself becomes non-optimal when the transmitter antenna count is less than the total receiver antennas [10]. Therefore, NOMA is necessary to provide additional power dimensions [11]. Thus, the researchers work towards MIMO-NOMA to overcome this issue.
In MIMO-NOMA, there are two different routes; namely, single-cluster, and multi-cluster [12]. In the single-cluster routes, all the network users other than the weakest one utilizes SIC [13]. But in multi-cluster MIMO-NOMA initially, all the users are partitioned into different clusters [14]. However, the entire benefit of MIMO-NOMA was attained by performing beam forming, power allocation, clustering, and SIC at the same time [15]. Mostly in the MIMO-NOMA technique, SIC decoding offers a perfect solution without errors in the channel [16]. However, the SIC-stability is one of the major challenges in NOMA-MIMO designs [17]. These stability issues often arise when achievable rates are applied in NOMA designs. The achievable rates-based NOMA technology causes a negative impact on SIC. For example, the error rate increases when equal power allocation is given to the design [18]. This may also lead to security issues, especially for the weakest user in the network. Moreover, many researchers were conducted to integrate the NOMA with millimeter wave MIMO to enhance the speed and network coverage [19]. However, massive antenna configuration is one of the major challenges in NOMA-mm Wave MIMO [20]. This challenge makes the conventional channel estimation techniques ineffective.
The other main concern in the traditional approach is the network coverage. Generally, a sufficient amount of power is required for data transmission in wireless networks. To improve the network coverage, an optimal range of power should be allocated to each users (far and near users). Normally, the edge user requires more power for data transmission. If more power is allocated for all users in the networks, it leads to high power consumption, and PAPR. Hence, improving the edge cell efficiency is important for optimizing the PAPR. Furthermore, some researchers investigate attaining an optimal channel estimation approach. Some existing methods like closed-form based Symbol Error Rate (SER) [21], Iterative Dichotomy for PAPR reduction [22], and multiple access approach named Index modulation multiple access (IMMA) [23], etc., are developed to overcome the problems with tradition channel estimation techniques. However, the high BER and PAPR reduce the channel performances. Thus, an optimal neural-based channel estimation method is developed in this article to improve the channel performances by minimizing the BER, and PAPR.
The key contributions of the presented model are listed as follows;
Initially, the transmitter parameters (NOMA-MINO) and input analog signal are initialized.
Convert the input signal into a digital format using ADC and then it is encoded into data streams.
Design the proposed WOA-RBFNN channel estimation model in the system to estimate the path.
Moreover, the WOA optimal fitness is incorporated into the neural network to tune the performance parameters.
Finally, the outcomes of the developed model are validated with the existing approaches in terms of Bit Error Rate (BER), throughput, network efficiency, etc.
The presented article is sequenced as follows; the recent works related to channel estimation in MIMO-NOMA are described in section 2, the existing system with its problem statement is detailed in section 3, the proposed model is explained with a flowchart and algorithm in section 4, the performances evaluation of the presented work is detailed in section 5, and the conclusion of the article is discussed in section 6.
Some of the recent literature related to the presented works are listed below,
Qinwei He et al [21] suggested a closed-form-based Symbol Error Rate (SER) derivative to estimate the reliability performances of the NOMA system. In this model, the reliability performance of the NOMA system with the Quadrature Amplitude Model (QAM) and Pulse Amplitude Modulation (PAM) are analyzed. Furthermore, the power allocation technique is introduced to decelerate the SER of the user. However, the suboptimal approach used in the model does not provide an effective solution.
In recent times, OFDM multicarrier modulation approach is used in wireless communication systems especially in 5th -generation systems for providing better spectral density. The major challenge in OFDM modulation is high PAPR which occurs when the nonlinear power amplifiers are operated near the saturation region. Thus, Khaled Tahkoubit et al [22] presented an efficient PAPR reduction model named Iterative Dichotomy. This method used analytical derivatives for decelerating the PAPR range. However, the implementation complexity is more.
In communication networks, NOMA is widely used to enhance network connectivity, and stability. However, the high error rate in NOMA demands an efficient channel estimation technique. Thus, Hamad Yahya et al [23] proposed a power-tolerant NOMA approach, which decelerates the network sensitivity by adaptively varying the signal power of the users. Although this scheme minimizes the error rate by 10dB, it is applicable for multicarrier modulation.
Saud Althunibat et al [24] presented an uplink multiple access approaches named Index modulation multiple access (IMMA). Here, each user's transmitted blocks are used to modulate a complex symbol, which is then transferred to a particular time slot. This presented model reduces the system complexity and error rate in the NOMA system. Furthermore, the designed model is validated with a comparative analysis. But, the probability of collision is high in this model.
Compared to OMA networks, NOMA provides greater spectral and network efficiency to multi-user communication systems. The existing techniques like Rayleigh fading channels cause low performances in the NOMA system. Therefore, Kuang-Hao Liu et al [25] developed a Quasi-degradation probability of 2-user NOMA over the Rician fading channel. This technique improves the system's performance and enhances network stability. However, the complexity and cost are high in this approach.
Generally, to minimize the MIMO dimension Single Beam selection approach is used in networks. But the user cannot properly select the multiple Radio Frequency (RF) chain, groups. Hence, Satyanarayana Murthy Nimmagadda [26] developed a hybrid based on the two optimization approaches named Alternative Grey Wolf with Beetle Swarm Optimization. This hybrid system acts as a multi-objective function to solve challenges regarding energy consumption, and power efficiency. However, it does not support future wireless communication systems.
Jawad Mirzaś et al [27] presented a two-stage channel estimation technique to overcome the issue with the traditional Reconfigurable Intelligent Surface (RIS) channels. The conventional RIS channels face ill-conditioned dictionary learning issues. The presented model utilizes a bilinear adaptive vector approximate message passing method to determine the RIS channels. This algorithm helps in identifying the ill-conditioned dictionary problems accurately. Moreover, the estimate of the phase shift matrix an approximate closed-form function is deployed in the system. However, the algorithm complexity and time consumption are more in this model.
In cellular networks, Intelligent Reflecting Surface (IRS) is widely used to improve the wireless propagation environment. However, the tuning of phase shifters in IRS is very complex. Thus, Tao Jiang et al [28] designed a machine-learning technique that tunes the beam formers optimally. This method deploys a deep neural network algorithm to tune the phase shifters optimally. In addition, a Graph neural network is utilized to capture the interrelations between the users. However, the error rate is more in this approach.
In 5th generation and beyond networks, IRS aided system is widely used to accelerate spectral power and energy efficiency. But the noise randomness is one of the major challenges in the IRS system. Therefore, Anastasios Papazafeiropoulos et al [29] developed a channel estimation method for reducing the computational complexity. The introduced technique includes closed-form derivative and low training overhead. Although the presented technique achieved better spectral efficiency, it does not allow for performing optimization which makes the IRS ineffective.
The major task in the IRS-aided communication system is finding the optimal channel estimation approach. Typically, the IRS system deploys a parallel channel with a sophisticated statistical distribution. The problem with the IRS communication system is a denoising issue. Hence, Chang Liu et al [30] designed a deep residual learning algorithm to retrieve the channel coefficients from the noisy parameters. Furthermore, the simulation outcomes of the developed model show better performances in minimizing the noise effects. Still, it does not provide the optimal solution for the denoising problem.
Recently, the NOMA approach is greatly applied in wireless communication systems to enhance the channel capacity and performance parameters. The basic principle behind NOMA is that it serves multiple users at the same time by performing power domain multiplexing and SIC at the transmitter and receiver side, respectively. Moreover, in communication systems, NOMA is combined with MIMO to attain high spectral efficiency. The MIMO is an antenna methodology, which widely used for minimizing multipath scattering and fading in a wireless communication system.
Thus, the combination of NOMA with MIMO provides high system performance in terms of network coverage and spectral efficiency. However, the high BER and less throughput are the major issues in the NOMA-MIMO system. Therefore, to overcome these demerits various channel estimation approaches are developed. However, the high BER in the channel degrades its network performance. Thus, estimating the path with less BER for data transmission is necessary. Moreover, to enhance the edge user capacity, optimizing the PAPR is important. Thus, an efficient channel estimation technique is required to overcome the challenges in the existing system. Hence, a novel optimization-based channel estimation approach is developed in this article to improve the system's performance. The system model and its problem statement are described in Fig. 1.
Nowadays, NOMA is widely applied in wireless cellular communication to offer better network connectivity, especially in 5G networks. It provides good stability and throughput compared to the OMA systems. Further, in the existing works, the NOMA is integrated with the MIMO in the communication channel to enhance spectral and power efficiency. However, the system performances are low due to high BER. Moreover, it demands an efficient channel estimation technique to increase the channel capacity of NOMA by accelerating the system performance in terms of BER. Thus, a novel hybrid technique named the WOA-ABFNN-based channel estimation approach is introduced in this article. The presented model incorporates the features of the whale optimization technique [31] and Radial basis neural network [32]. In the initial phase, the transmitter parameters and input data are initialized in the system. Then, the input analog data is converted into digital form using the Analog-to-Digital (ADC) converter.
Further, using the channel encoder and modulator the input data is encoded into data streams. Then, the proposed method is developed in the system with a fitness function. Here, the fitness function of the WOA approach is incorporated in the hidden layer of the RBFNN to estimate the optimal path for data transmission in the communication channel. Moreover, it provides the finest outcomes by tuning the channel parameters. The presented model is implemented in the MATLAB environment and the results are estimated. Furthermore, the estimated outcomes are evaluated by comparing them with the performances of the existing channel estimation approaches. The proposed methodology is described in Fig. 2.
Consider a MIMO system containing a base system with \({N}_{TAc}\)antennas and \({N}_{TAc}^{rf}\) RF chains that transmits the input signal to a single user with \({N}_{RAc}\)receiver antennas and \({N}_{RAc}^{rf}\) RF chains. In base stations, the phase shifters are used to connect more antennas with few RF chains. Hence, the transmitter antennas \({N}_{TAc}\)and receiver antennas\({N}_{RAc}\) should be highly greater than both transmitter and receiver RF chains (\({N}_{TAc}^{rf}\)and\({N}_{RAc}^{rf}\)), respectively. In the delay domain, the channel matrix (\({N}_{TAc}\times {N}_{RAc}\)) between the transmitter and receiver is expressed in Eq. (1).
Where, \(\tilde{L}\)represents the main path count, \({\alpha }_{n}\)denotes the propagation gain of the nth path, \({\stackrel{̑}{T}}_{n}\)indicates the nth path delay, \({\varphi }_{n}\)and \({\phi }_{n}\)refers to the azimuth angles of arrival and departure at transmitter and receiver, respectively. The response vector for the uniform linear array is represented in Eqs. (2) and (3).
\({{\alpha }_{R}}^{{\prime }}\left(\varphi \right)=\frac{1}{\sqrt{{N}_{RAc}}}{\left[1,{e}^{-j2\pi \frac{\kappa }{{\lambda }^{{\prime }}}{sin}{\varphi }_{n}},....,{e}^{-j2\pi \frac{\kappa }{{\lambda }^{{\prime }}}\left({N}_{RAc}-1\right){sin}{\varphi }_{n}}\right]}^{\stackrel{̑}{T}}\) (2) \({{\alpha }_{T}}^{{\prime }}\left(\varphi \right)=\frac{1}{\sqrt{{N}_{TAc}}}{\left[1,{e}^{-j2\pi \frac{\kappa }{{\lambda }^{{\prime }}}{sin}{\phi }_{n}},....,{e}^{-j2\pi \frac{\kappa }{{\lambda }^{{\prime }}}\left({N}_{TAc}-1\right){sin}{\phi }_{n}}\right]}^{\stackrel{̑}{T}}\) (3)
Where \(\kappa\)refers to the distance between the nearby antennas, and \({\lambda }^{{\prime }}\)indicates the carrier wavelength.
Generally, in NOMA designs, different powers are allocated to multiple nodes (users) at the common orthogonal block. Therefore, fewer power levels are allocated to nodes with good channel circumstances. For example, consider two users i.e., user 1 and user 2, and assume that the channel condition is good for user 1 when compared to user 2. Hence, the power allocation for user 1 is less compared to user 2. In the receiver end, user 1 first decodes the input data considering user 2 as external noise. Hence, direct detection is not feasible for user 1 since more power is allocated to user 2. User 1 decodes and detects user 2 signal first and to cancel that it uses SIC. Then, user 1 retrieves its own signal from the noise.
Dynamic clustering utilizes gain variance among the multiple users in the NOMA cluster and then joins them into single or multiple clusters. This helps in improving the throughput of the system. In download link NOMA, it pairs the first lowest and highest gain users into the common NOMA cluster. Similarly, the second lowest and highest gain users are paired into another cluster, and so on. This process increases the throughput of the system by pairing the highest and lowest gain users. Here, \({N}_{RAc}\)users are grouped into the same NOMA cluster\({N}_{cl}\) with each cluster having two gain users (highest and lowest gain users). In each cluster, the channel correlated with users is grouped into a channel matrix. It is expressed in Eq. (4).
Where,\({{H}^{*}}_{Ch}\in {C}^{{N}_{RAc}\times {N}_{TAc}}\) and\({N}_{cl}\) contains both weak and strong channel users.
Beamforming is performed to alleviate the inter-cluster interference in the NOMA system. It reduces the inter-cluster interference by maximizing the signal towards the actual cluster and places null values in the interference direction. In Beamforming, beams from one cluster are made orthogonal to the users from another cluster. Consider\({h}_{i}^{{\prime }}\) and\({h}_{j}^{{\prime }}\) as strong and weal user in cluster\({k}_{c}\), where\({{H}^{*}}_{{k}_{c}}={\left[{h}_{i}^{{\prime }}{h}_{j}^{{\prime }}\right]}^{{T}^{{\prime }}}\). On the other hand, the channel coefficients of users in the cluster\({k}_{c}\) is expressed in Eq. (5).
\({{H}^{*}}_{-{k}_{c}}={\left[{H}_{1}^{{\prime }}{H}_{2}^{{\prime }}........{H}_{{k}_{c}-1}^{{\prime }}{H}_{{k}_{c}+1}^{{\prime }}.....{H}_{{N}_{Cl}}\right]}^{{T}^{{\prime }}}\in {C}^{\left({N}_{RAc-2}\right)\times {N}_{TAc}}\) (5) \({B}_{{k}_{c}}^{{\prime \prime }}={I}_{{N}_{TAc}}^{{\prime }}-{{H}^{{\prime }}}_{-{k}_{c}}^{H}{\left({H}_{-{k}_{c}}^{{\prime }}{{H}^{{\prime }}}_{-{k}_{c}}^{H}\right)}^{-1}{H}_{-{k}_{c}}^{{\prime }}\) (6)
The beamforming \({{H}^{*}}_{-{k}_{c}}\)is represented in Eq. (6). This beamforming matrix \({B}_{{k}_{c}}^{{\prime \prime }}\)will be orthogonal to channels of users in another cluster.
The presented model uses the features of RBFNN, and WOA to accelerate the channel capacity in the MIMO-NOMA system. RBN is a type of artificial neural network which has high learning speed. RBN is different from other NN approaches because of its approximation function. In the presented model, the utilization of RBN reduces the computational time and helps in estimating the appropriate path for data transmission. The path estimation is formulated in Eq. (7) \({E}_{s}\left(P\right)={exp}\left(-\frac{{‖{S}_{NR}-{E}_{R}‖}^{2}}{C{{h}_{c}}^{2}}\right)\) (7)
Where, \({E}_{s}\)indicates the path estimation variable, \(P\)refers to the path for data transmission, \({S}_{NR}\)represents the SNR, \({E}_{R}\)denotes the error rate, and\(C{h}_{c}\) indicates the channel capacity. Further, the channel parameters are tuned using the fitness function to achieve the finest channel performance. To tune the parameters, the optimal fitness function of WOA is integrated into the hidden layer of the RBFNN. The WOA is a nature-inspired meta-heuristic algorithm, which is used to solve different optimization problems. WOA is based on the behaviour characteristics of the humpback whale. One of the optimal features of the humpback whale is it can track the location of prey and encircle them. Here, it tracks the edge users and near users effectively and then it optimizes the power allocated to each user in accordance with their power requirement. Thus, it improves the edge user capacity by reducing the PAPR. The optimal tuning is expressed in Eq. (8).
Where \({F}_{Woa}^{{\prime }}\)represents the WOA fitness function, \({O}_{pv}^{{\prime \prime }}\)indicates the optimal value of the channel parameters, \({C}_{hl}^{{\prime \prime }}\)refers to the updated fitness value, and \({C}_{hl}\)denotes the channel parameters. Here, the fitness function tunes the channel performance parameters and updates them.
The flowchart of the proposed WOA-RBFNN model is illustrated in Fig. 3. Thus, the proposed model increases channel performance by reducing the BER, and PAPR.
A hybrid optimized channel estimation technique is developed in this article to increase the performance parameters by minimizing the BER, PAPR, and computational complexity. The presented model is executed in MATLAB software version R2021a. Initially, the transmitter and receiver parameters are initialized in the system. The proposed model initially estimates the path for data transmission by predicting the BER.
Table.1 Implementation Parameters
|
Parameters |
Description |
|---|---|
|
MATLAB |
R2021a |
|
OS |
Windows 10 |
|
Antenna |
4x4 |
|
Subcarriers |
64 |
Further, the optimal fitness function is integrated to tune the channel performance parameters. Finally, the system performances are evaluated and then compared with the existing purposes for validation purposes. The parameter and its description are listed in Table 1.
Case Study:
The working process of the proposed model is explained with a case study. Initially, a MIMO-NOMA system was built in the MATLAB tool with required number of users. Then, the proposed deep learning-based scheme was introduced to minimize the PAPR and improve the network performances. The main objective of the proposed model is to estimate the optimal path for data transmission. Initially, the RBFNN-WOA model is trained to find the optimal path and then the fitness function is deployed to increase the network performances. The network performance is improved by minimizing the PAPR and BER. Generally, in wireless networks there are two types of users namely, edge users and near users. The users nearer to the base station is near user and the users away from the base station are edge users. The near user requires less power for data transmission, while the edge user requires high power. But in the traditional schemes same amount of power is allocated for all types of users. Hence, the power usage, and energy waste is high. Thus, the network performances are low in the conventional approaches. This drawback is overcoming in the proposed model by allocating less power to the near user and optimal power to the edge user. For example, consider two users namely, user1 and user2. User1 is nearer to the base station and the user2 is far away from the base station. In the conventional technique, if 5W power is allocated for near user (user1) then the same power is allocated for the edge user (user2). However, the user1 requires less power for data transmission. This increases the PAPR of the system. But in the proposed model, it is minimized by reducing the power offered to the near user (user1) and compensatively offers the required power to the edge use (user2). This reduces the PAPR of the system and improves the throughout and network efficiency. The optimal fitness solution in the proposed model effectively allocates the required power to each user in the network. Thus, it tunes the channel parameters and improves the edge cell capacity.
In performance assessment, the network performance parameters like BER, throughput, network efficiency, and PAPR are estimated by implementing the proposed channel estimation model in MATLAB software.
In a communication network, BER indicates the rate of bits with errors corresponding to the sum of actual bits received during the data transmission. Generally, for a stable communication network, the BER should be low. The BER is expressed in Eq. (9).
Where, \({B}_{ER}\)indicates the BER, \({T}_{br}\)represents the sum of bits received with error, and \({T}_{bt}\)denotes the total number of bits transferred.
The BER of the proposed model is estimated by changing the antenna count to N = 2, N = 3, and N = 4. Figure 4 displays the BER of the developed model for different antenna sizes. Here, the BER is determined by adjusting the SNR arrangement from 0 to 30 dB. When N = 4, the BER of the developed model is\(5.3\times 1{0}^{-7}\) at 30 dB. Similarly, at 30dB SNR, the developed model attained BER of \(6\times 1{0}^{-5}\)and\(2\times 1{0}^{-5}\) for N = 3, and N = 2, respectively. From the observation, it is noticed that when antenna count (N) increases the BER decreases. Thus, the presented channel estimation enhances the channel performances by minimizing the BER.
Normalized Error Square Error (NMSE) is the MSE normalized by the signal power. It is predicted by dividing the square of the difference between the desired and estimated value by the square of the desired value. It is represented in Eq. (10).
Where \({N}_{MSE}\)represents the NMSE of the presented model, \({D}_{V}\)refers to the desired outcome value, and \({E}_{V}\)indicates the estimated outcome.
The robustness and performance of the designed model are weighed by calculating the NMSE of the NOMA-MIMO channel. Here, the NMSE is determined by SNR. The developed model attained NMSE of\(6\times 1{0}^{-3}\),\(2\times 1{0}^{-3}\) and\(4\times 1{0}^{-5}\), respectively for N = 2, N = 3, and N = 4. The NMSE performance of the developed model is displayed in Fig. 5. The NMSE analysis shows that in the presented model the NMSE decreases with an increase in antenna count (N).
PAPR represents the ratio of the peak to the average power of a signal. Generally, this is estimated for a transmitted signal in an OFDM system and it is expressed in decibels (dB). The lower PAPR value indicates the stable and efficient performance of the communication system. The PAPR of the communication system is formulated in Eq. (10).
Where, \({P}_{APR}\)indicates the PAPR value of the communication system, \({H}_{PP}\)indicates the highest peak power and \({T}_{AP}\)denotes the total average power.
Typically, the PAPR is estimated in accordance with the complementary cumulative distribution function (CCDF). The PAPR attained by the presented model is 2 dB. The PAPR performance of the designed model is displayed in Fig. 6. In the developed scheme, the PAPR is minimized by reducing the power allocation to the near user.
The presented model is validated by comparing the performance parameters like BER, throughput, PAPR, and network efficiency with other existing techniques. The existing methods used for comparative analysis are Compressive Sampling Matching Pursuit Greedy Approach (CSMPG) [33], Noise Corrected-Clustered Space Channel Estimation (NC-CSCE) [34], Faster-than Nyquist Signaling (FTN-NOMA) [35], Deep Learning-based Detector (DLD) [36], Sparity Adaptive Matching Pursuit (SaMP) [37], Dynamic Step size- Sparity Adaptive Matching Pursuit (DSS-SaMP) [37], Adaptive Step size- Sparity Adaptive Matching Pursuit (AS-SaMP) [37], and Nonlinear Distortion Removal-based on Deep Neural Network (NDR_DNN) [38], and Fully Connected Deep Neural Network (FC_DNN) [39].
The robustness of the developed model is validated by equating the BER with the other conventional techniques. Here, it is compared with channel estimation techniques like CSMPG, NC-CSCE, and FTN-NOMA.
At 30dB SNR, the proposed model attained less BER of\(5.3\times 1{0}^{-7}\), whereas the existing techniques earned high BER of\(1\times 1{0}^{-3}\),\(56\times 1{0}^{-4}\) and\(83\times 1{0}^{-5}\), respectively. The BER comparison is illustrated in Fig. 7. The less BER verifies that the channel performance is high in the presented model.
PAPR is one of the important factors which determine the performance of the communication channel. Generally, the low PAPR indicates better channel performance. The PAPR performance of the developed model is validated with a comparative analysis.
Here, the PAPR of the designed model is compared with the existing techniques like FC_DNN, SaMP, and NDR_DNN. The PAPR value earned by the existing methods is 10.5dB, 8.5dB, and 7.8dB, respectively. But the presented hybrid model achieved less PAPR of 2dB. This shows that the PAPR is low in the proposed model. The comparison of PAPR is displayed in Fig. 8.
NMSE defines the difference between the actual and desired signal outcomes. For an effective communication channel, the NMSE value should be low. The NMSE of the developed channel estimation approach is verified by comparing it with the conventional approaches.
The NMSE earned by the recommended model is equated with the other systems like CSMPG, NC_CSCE, and FTM_NOMA. The NMSE attained by the existing methods is\(3\times 1{0}^{-2}\), \(7.4\times 1{0}^{-4}\) and\(4.2\times 1{0}^{-4}\), respectively. The NMSE of the proposed model at 30dB SMR is\(4\times 1{0}^{-5}\). This comparative analysis shows that the NMSE attained by the designed model is low compared with the existing models. The NMSE validation is shown in Fig. 9.
Computational complexity represents the amount of computing resources (space and time), a system consumes for a particular algorithm. To manifest that the developed model consumed less time and space, the computational time is compared with existing approaches like SaMP, DSS_SaMP, and AS_SaMP.
Table .2 Computational Time Comparison
|
Methods |
Computational Time (s) |
|---|---|
|
SaMP |
0.0016 |
|
DSS_SaMP |
0.0011 |
|
AS_SaMP |
0.0013 |
|
WOA_RBFNN |
0.00074 |
The computational complexity attained by the existing techniques is 0.0016s, 0.0011s, and 0.0013s, respectively and the time taken by the presented model is 0.00074s. The comparison of the running time is tabulated in Table 1. Thus, the comparative assessment of computational time shows that the complexity of the developed model is less when compared to the existing channel estimation model.
A novel channel estimation technique is designed in this article to improve channel performance metrics by estimating the path for data transmission. In the proposed scheme, the WOA fitness function is incorporated in the hidden layer of the RBFNN to determine the optimal path for data transmission. Moreover, it helps in tuning the channel parameters like BER, PAPR, and NMSE till it reaches their optimal value.
|
Metrics |
Performances (At SNR = 30dB) |
||
|---|---|---|---|
|
N = 2 |
N = 3 |
N = 4 |
|
|
BER |
\(2\times 1{0}^{-5}\) |
\(6\times 1{0}^{-5}\) |
\(5.3\times 1{0}^{-7}\) |
|
NMSE |
\(6\times 1{0}^{-3}\) |
\(2\times 1{0}^{-3}\) |
\(4\times 1{0}^{-5}\) |
The performance assessment is listed in Table 3. When SNR = 30dB, the presented hybrid model earned less BER of\(2\times 1{0}^{-5}\), \(6\times 1{0}^{-5}\) and\(5.3\times 1{0}^{-7}\), respectively for N = 2, 3, and 4. Similarly, the NMSE achieved by the designed model is\(6\times 1{0}^{-3}\),\(2\times 1{0}^{-3}\) and\(4\times 1{0}^{-5}\) for N = 2, 3, and 4, respectively. Furthermore, the estimated performances are validated with a comparative analysis. The comparative analysis verifies that the presented model earned better performances than the existing techniques.
Generally, in MIMO-NOMA the channel performance is degraded due to the SIC issue. Hence, a novel optimized neural approach is proposed in this paper to increase the channel capacity and its performance. Initially, the MIMO-NOMA system was designed system with transmitter and receiver parameters. The proposed channel estimation model is integrated into the transmitter phase to determine the path for data transmission. Further, the channel performance parameters are tuned using the WOA fitness function. In addition, the proposed scheme effectively reduces the PAPR and enhances the system performances. The optimal fitness function helps to attain the finest channel performance. Finally, the outcomes are estimated and the parameter enhancement score is determined from the comparative analysis. The comparative analysis shows that the BER is reduced by\(7.7\times 1{0}^{-5}\), the PAPR is minimized by 5.8dB, and NMSE is decreased by\(2.9\times 1{0}^{-2}\). The comparative analysis proved that the presented model earned better performance than others.
Ethical approval: My research guide reviewed and ethically approved this manuscript for publishing in this Journal.
Author Contributions statement: The authors confirm contribution to the paper as follows: Study conception and design Jenish Dev. M, Judson. D; Data collection: Jenish Dev. M; Analysis and interpretation of results: Judson. D; Draft manuscript preparation: Jenish Dev. M, Judson. D; All authors reviewed the results and approved the final version of the manuscript.
Competing interests: This paper has no conflict of interest for publishing
Research funding: No Financial support
Availability of data and material: Data sharing is not applicable to this article as no new data were created or analyzed in this Research.
Human and Animal Rights: This article does not contain any studies with human or animal subjects performed by any of the authors.
Informed consent: I certify that I have explained the nature and purpose of this study to the above-named individual, and I have discussed the potential benefits of this study participation. The questions the individual had about this study have been answered, and we will always be available to address future questions.
Acknowledgements: The authors would like to thank the reviewers for all of their careful, constructive and insightful comments in relation to this work.