A low complexity enhanced-NOMA scheme to reduce inter-user interference, BER and PAPR in 5G wireless systems

Abstract

Non-orthogonal multiple access (NOMA) has been recognized as an attractive technique to address the increasing traffic demands for heterogeneous wireless networks. The conventional fast Fourier transform NOMA (FFT-NOMA) system causes inter-user-interference (IUI), and high peak to average power ratio (PAPR) because of the superposition of multiple users on the same frequency subchannel. To reduce the PAPR, IUI, and bit error rate (BER) in a FFT-NOMA system, the enhanced-NOMA (E-NOMA) scheme is proposed that consists of a new low-complexity modified version of the conventional selected mapping (CSLM) cascaded with the Walsh–Hadamard transform (WHT). Simulation results demonstrate that the proposed E-NOMA scheme outperforms the FFT-NOMA with reference to PAPR and BER about 4.3 dB and 9.5 dB, respectively. Furthermore, at least 56% computational complexity reduction is achieved compared to an SLM-based NOMA method.

Introduction

With the exponential increment of mobile data traffic, new multiple access techniques have been developed to meet the demand of high spectral and energy efficiencies, low latency, and proper quality of service (QoS) in $5$G wireless communication systems [1]. Since the resource management in the orthogonal frequency division multiple access (OFDMA) network is based on the frequency allocation, where each resource block is allocated to a single user at each time slot, then the QoS and the interference management for a large number of users are the main challenges of the utilization of OFDMA in the $5$G networks [2], [3]. Among several approaches to meet the expectations of $5$G networks and beyond, non-orthogonal multiple access (NOMA) attracts more attention of researchers in recent years [4]. The NOMA technique shares the same frequency channel between multiple users simultaneously, offering several benefits such as spectrum efficiency and fairness among base station (BS)-close and edge users [5]. To be more precise, NOMA solves the frequency limitation of OFDMA networks using this new resource allocation strategy. Generally, NOMA schemes are broadly categorized in code-domain NOMA (C-NOMA) and power-domain NOMA (P-NOMA) classes [6]. The P-NOMA scheme schedules the power allocation of users, based on the channel state in terms of their bandwidth, fading channel gain, noise power, and the transmission data rate. Furthermore, it considers an individual weighted coefficient factor for each user to allocate the same subchannel to multiple users. Consequently, the power distribution of the P-NOMA scheme is accomplished by the channel state information (CSI) and an individual factor, simultaneously [7].

In the NOMA architecture, superposition-coding (SC) at the transmitter and successive interference cancellation (SIC) [8] at the receiver are employed to send different signals to several receivers simultaneously, and to separate the signals of multiple users, respectively. Since P-NOMA distinguishes user entities by transmitting powers in the same subchannel, inter-user-interference (IUI) increases the bit error rate (BER), significantly [9]. Hence, despite several benefits in deploying NOMA in various wireless networks, the SIC procedure is one of the substantial challenges in the practical implementation of the NOMA technique in current wireless systems. This comes from the decoding complication and the error propagation in the NOMA system [10]. Consequently, due to the error growth in proportion to the users’ increment, traditional NOMA without any SIC-enhancement technique, would not be an admirable solution to use a shared frequency subchannel. This issue would be more critical in the next generation of wireless networks which have been recognized as ultra-dense heterogeneous networks, where lots of smart devices in dense areas are assigned NOMA techniques [11]. Moreover, $5$G networks are planned to use millimeter-wave frequency ranges. This strategy leads to higher levels of out-of-band distortion at the non-linear region of the power amplifiers (PAs) [12], compared to the previous generation. To improve the spectrum efficiency, the orthogonal frequency division multiplexing (OFDM)-based NOMA system [13] has attracted a lot of attention compared to other traditional NOMA techniques. Preventing the signal to pass through the non-linear region in OFDM-based NOMA systems needs a low peak to average power ratio (PAPR) value. Therefore, efficient low-complexity PAPR minimization schemes are yet crucial in such systems.

To tackle the aforementioned problems associated with NOMA, recently, several promising approaches have been developed. In [13], a low complexity algorithm is proposed to jointly optimize subcarrier assignment and power allocation with the aim of minimizing the total transmission power in OFDM-based NOMA systems. However, this method suffers from a high PAPR and high complexity due to the inverse fast Fourier transform (IFFT) [14] implementation. The sparse code multiple access (SCMA)-based NOMA scheme in [15] uses multi-dimensional constellations to provide a statistical dependency between data-carrying subcarriers. Since the typical SCMA constellation design in this approach does not affect the BER performance, the conventional selected mapping (CSLM) without any BER reduction block can be applied to such systems. CSLM multiplies the input symbols by the random phase-selections among $\{1,-1,j,-j\}$, with the aim of finding the optimum phase sequence, creating the minimum PAPR value [16]. However, due to the random phase sequence property of CSLM, it yields highly redundant computational complexity. A low computational complexity scheme for the PAPR reduction is presented in [17] that employs linear scaling of a portion of signal coefficients by an optimal factor. Despite the acceptable complexity computation of the scheme in [17], the achieved PAPR value and the target rate cannot satisfy $5$G communication expectations.

A hybrid PAPR reduction scheme is proposed in [18] to minimize the PAPR of a $5$G system. The approach divides the whole frequency band into a number of sub-channels, then the data stream of each sub-channel is processed by an IFFT and a filter in series. However, this scheme imposes high computational complexity on the network due to a large number of IFFT blocks. The same drawback can be tracked at [19] where a multi numerology OFDM scheme with new subcarrier spacing is proposed. In [20] , frequency spreading filter bank multicarrier (FS-FBMC) is introduced as a multicarrier candidate for $5$G wireless communication systems. To mitigate the inter-carrier interference (ICI) and the inter-symbol interference (ISI), both FS-FBMC and OFDM are combined, while the latter resources are reused for FS-FBMC. Despite the great achievement of this combination in terms of interference mitigation, a large number of IFFT blocks and the presence of a filter bank cause the computational complexity increment. Since OFDM continues to be used in $5$G and is likely to be studied in beyond $5$G communication systems as well, [21] proposes a regularization optimization-based flexible hybrid companding and clipping scheme (ROFHCC), to reduce the PAPR of OFDM systems. Since, companding and clipping are not basically distortion-less methods, the ROFHCC scheme suffers from high distortion and non-linearity characteristics at the high numbers of users in real $5$G systems. Also, the same drawback has been tracked in [22] when clipping noise is separated from the signal, and tailored frequency-selective clipping noise filtering is used to control the tradeoff between PAPR value and the quality of the transmitted signal. Authors in [23] apply a discrete-cosine transform matrix (DCTM) precoding to the constellation symbols to decrease the autocorrelation ratio among modulated data. The main drawback of this method is the high value of IUI which degrades the overall system’s efficiency. A centralized power allocation approach based on a gradient descent algorithm is proposed in [24] that needs a large number of iterations in users’ increment. Some recent literature proposes various schemes to eliminate the IFFT blocks. For instance, [25] uses discrete wavelet transform (DWT) to enhance the spectral efficiency and the capacity of the NOMA-based wireless network. This idea uses the fact that a signal is essentially fragmented into its low- and high-frequency components through wavelet filters. However, due to having a unique center frequency, a large delay would appear in such a system. Ref. [26] combines generalized frequency division multiplexing (GFDM) with NOMA schemes to extend the framework of multi-carrier NOMA networks. Such a NOMA joined GFDM modulator per user displays a high computational complexity without a noticeable PAPR reduction value when compared to the FFT-NOMA method.

In this paper, we propose a new method, named enhanced-NOMA (E-NOMA), to efficiently decrease the IUI, BER, PAPR, and the computational complexity in $5$G networks, compared to the FFT-NOMA and some recent related methods. Ignoring any of the aforementioned metrics leads to efficiency degradation. The proposed scheme is based on cascading a new low-complexity modified version of the CSLM method with the Walsh–Hadamard transform (WHT) utilization approach. The first contribution of the E-NOMA approach is to mitigate the co-channel and inter-symbol interferences. This is accomplished by heuristically employing the Walsh–Hadamard matrix to allocate the NOMA symbols orthogonally. It is demonstrated that the orthogonality of the user streams at each sub-frequency-channel and orthogonality of sub-channels significantly mitigate the IUI and inter-symbol interference (ISI), respectively. In this case, the efficiency of the SIC system can be improved using predetermined allocating orthogonal data streams by the Walsh–Hadamard transform (WHT). Since the orthogonality property of Hadamard matrices may get lost in the presence of phase-shifting [27], employing the SIC technique is yet needed at the receiver. We then propose a new low-complexity modified version of CSLM that offers the PAPR and computational complexity reduction compared to the traditional OFDM-based NOMA systems. The analysis supported by simulation results shows that the proposed E-NOMA scheme is superior to the FFT-NOMA in terms of the PAPR and BER reduction of about 4.3 dB and 9.5 dB, respectively. In addition, at least $56$% of computational complexity reduction is achieved through deploying our scheme compared to the SLM-based NOMA method.

The remainder of the paper is organized as follows: Section 2 presents some preliminary definitions including the PAPR and WHT concepts. The proposed E-NOMA scheme is introduced in Section 3. The computational complexity analysis is presented in Section 4. Section 5 deals with discussing on the simulation results in terms of PAPR and BER, and is followed by the conclusion in Section 6.

Notation

Throughout the paper, we use boldface letters to denote vectors and matrices. Operations $\otimes$, $\mathbb{E}[.]$ and $\circ$ denote the Kronecker product, expectation operator, and the element-wise multiplication, respectively.