The massive surge in device connectivity demands higher data rates, increased capacity with low latency and high throughput. Hence, to provide ultra-reliable, low-latency communication with ubiquitous connectivity for Internet-of-Things (IoT) devices, next-generation wireless communication leverages the incorporation of machine learning tools. However, standard data-driven models often need large datasets and lack interpretability. To overcome this, model-driven deep learning (DL) approaches combine domain knowledge with learning to improve accuracy and efficiency. Deep unfolding is a model-driven method that turns iterative algorithms into deep neural network (DNN) layers. It keeps the structure of traditional algorithms while allowing end-to-end learning. This makes deep unfolding both interpretable and effective for solving complex signal processing problems in wireless systems. We first present a brief overview of the general architecture of deep unfolding to provide a solid foundation. We also provide an example to outline the steps involved in unfolding a conventional iterative algorithm. We then explore the application of deep unfolding in key areas, including signal detection, channel estimation, beamforming design, decoding for error-correcting codes, integrated sensing and communication, power allocation, and physical layer security. Each section focuses on a specific task, highlighting its significance in emerging 6G technologies and reviewing recent advancements in deep unfolding-based solutions. Finally, we discuss the challenges associated with developing deep unfolding techniques and propose potential improvements to enhance their applicability across diverse wireless communication scenarios.

This paper proposes, for the first time, a hybrid multiple access framework that integrates the principles of rate-splitting (RS) and sparse code multiple access (SCMA) in an SISO downlink scenario. The proposed scheme, termed RS-SCMA, unifies the powerful interference management capability of rate-splitting multiple access (RSMA) with the near-optimal multiuser detection of SCMA. A key feature of RS-SCMA is a tunable splitting factor α, which governs the allocation between the genericM-ary modulated common messages and SCMA-encoded private messages. This enables dynamic control over the fundamental trade-off between system sum-rate, bit error rate (BER), and the overloading factor. We develop novel transmitter and receiver architectures based on soft successive interference cancellation (SIC), incorporating message passing algorithm (MPA) detection and soft-symbol reconstruction. Furthermore, a unified analytical expression for the achievable sum-rate is derived as a function of the splitting factor α. The performance of the proposed RS-SCMA system is evaluated in terms of both BER and sum-rate. Simulation results confirm the superiority of RS-SCMA over conventional SCMA and multi-carrier RSMA, demonstrating its scalability and robustness even in the presence of channel estimation errors.

Visual question answering (VQA) poses a multi-modal translation challenge that requires the analysis of both images and questions simultaneously to generate appropriate responses. Although VQA research has mainly focused on text-based questions in English, speech-based questions in English and other languages remain largely unexplored. Incorporating speech could significantly enhance the utility of VQA systems, as speech is the primary mode of human communication. To address this gap, this work implements a speech-based VQA system and introduces the textless multilingual visual question answering (TM-VQA) dataset, featuring speech-based questions in English, German, Spanish, and French. This TM-VQA dataset contains 658,111 pairs of speech-based questions and answers based on 123,287 images. Finally, a novel, cross-attention-based unified multi-modal framework is presented to evaluate the efficacy of the TM-VQA dataset. The experimental results indicate the effectiveness of the proposed unified approach over the cascaded framework for both text and speech-based VQA systems. Dataset can be accessed at https://github.com/Synaptic-Coder/TM-VQA.

Even though large language models (LLMs) have demonstrated remarkable performance across various natural language processing tasks, their application in speech-related tasks has largely remained underexplored. This work addresses this gap by incorporating acoustic features into an LLM which can be fine-tuned for downstream direct speech-to-text translation and automatic speech recognition tasks. To address the computational demands associated with fine-tuning LLMs, a novel self and semi-supervised knowledge distillation technique is proposed to implement a lightweight LLM having 50% lesser parameters. Validated on the MuST-C and Librispeech datasets, this technique achieves over 92% of the performance of the larger LLM, demonstrating both robust performance and computational efficiency.

Intelligent reflecting surfaces (IRS) have emerged as a transformative technology for enhancing wireless communication by dynamically reconfiguring the propagation environment. Additionally, orthogonal time frequency space (OTFS) has been recognized as a promising waveform for high-mobility scenarios. In this work, we investigate the joint optimization of beamforming at the base station (BS) and phase shifts at the IRS in an IRS-aided multiple-input single-output (MISO) OTFS system. The objective is to maximize the achievable sum rate under practical constraints, resulting in a highly non-convex optimization problem involving beamforming and phase shifts. To address this challenge, we propose a Bayesian neural network (BNN) based approach that models uncertainties in channel state information (CSI) and optimally learns both the beam-forming vector and IRS phase shift matrix. The BNN employs a probabilistic weight distribution, enabling robust optimization through stochastic gradient-based updates. Furthermore, we leverage Monte Carlo sampling to account for uncertainties in the learned parameters, ensuring improved performance and robustness under dynamic channel conditions. Simulation results demonstrate that the proposed BNN-based optimization outperforms conventional methods, such as alternate optimization (AO), by achieving higher capacity. Additionally, its probabilistic framework enables robustness to variations in CSI, making it well-suited for dynamic channel conditions.

Orthogonal time-frequency space (OTFS) modulation is a promising technique for next-generation wireless communication, particularly in high-mobility scenarios. However, the performance of the OTFS modulation is degraded by hardware impairments (HIs), particularly at high frequencies and mobility. Thus, efficient signal detection is essential for reliable OTFS-based communication. To overcome this, various detectors, including convolutional neural networks (CNN), minimum mean square error (MMSE), and message passing (MP), have been explored in the literature, with CNN-based detectors demonstrating superior performance over MMSE and MP. In this work, we propose a residual channel attention network (RCAN) to robustly detect OTFS signal in the presence of HIs. Using deep residual learning and channel attention mechanisms to improve detection accuracy under varying channel conditions. Extensive simulations, considering different OTFS frame sizes, user mobility levels, and HI parameters, confirm that the proposed RCAN consistently outperforms conventional detectors, including CNN, in all scenarios evaluated. This makes RCAN a highly effective solution for OTFS-based high-mobility wireless communication systems, both with and without HI.

Terahertz (THz) wireless communication is a promising technique to meet the rising demand for high-bandwidth services. Propagation of the $\mathbf{T H z}$ waves through the atmosphere is influenced by factors like attenuation, water vapour, weather, turbulence, rain and beam misalignment. Furthermore, performance of a THz link is severely degraded by non-linear hardware impairments introduced by power amplifier (PA). Conventional estimators such as zero-forcing (ZF) and minimum mean squared error (MMSE) fall short of handling beam misalignment, rain attenuation, and non-linearity. To circumvent this limitation, we propose a deep neural network (DNN) based receiver for THz communication systems in the presence of PA non-linearity. Simulation results for bit error rate (BER) performance considering non-linear distortion due to power amplifier indicate that the proposed deep learning (DL) based receiver performs better in terms of BER and is more robust compared to ZF, MMSE and orthogonal matching pursuit (OMP) based receiver.

Efficient data transmission is one of the most important features that any technology should possess. However, this efficiency should be achieved within the available resources, which in this context is energy. One method to achieve it is the use of hexagonal quadrature amplitude modulation (H-QAM). The recently proposed Orthogonal Time Frequency Space (OTFS) modulation is considered for high mobility communication scenarios over Orthogonal Frequency Division Multiplexing (OFDM) due to its unique multiplexing capabilities in the delay-Doppler (DD) domain. Nevertheless, there is still room for advancement in OTFS system design. In this paper, our objective is to enhance the coding gain and shaping gain of OTFS systems, which can complement the existing advantages of OTFS technology. To achieve this, we explore square quadrature amplitude modulation (S-QAM) and H-QAM models and calculate and compare their shaping and coding gains, supported by theoretical analysis. Our simulations demonstrate that H-QAM outperforms S-QAM, primarily because H-QAM provides greater tolerance to noise by expanding the Voronoi region. We analyze the impact of number of delay and Doppler bins on the performance of the OTFS for various modulation orders. Additionally we also find the closed form of bit error rate (BER) by considering the nature of the eigenvalues of the channel matrix. A comparison of peak-to-average power ratio (PAPR) is also done.

Orthogonal time frequency space (OTFS) modulation is positioned to be a potential waveform for 6G communications, owing to its capability to multiplex data in the delay-Doppler (DD) domain, making it robust against doubly-selective wireless communications channels. OTFS leverages a 2D transformation that effectively transform a doubly-dispersive channel into one with minimal fading characteristics. In this paper, we develop a deep learning-based signal detector for a MIMO-based OTFS system (termed as DL-MOTFS) with hardware impairments (HIs). We assess the performance of the proposed DL-MOTFS for a range of system parameters, including those related to HIs, user velocity, OTFS frame size, and the number of transmit and receive antennas. Through comprehensive comparisons against the conventional MMSE (minimum mean square error) signal detector, we find that the proposed DL-MOTFS consistently outperforms the MMSE detector, exhibiting a performance gain of around 3–5 dB.

Orthogonal time frequency space (OTFS) has emerged as the most sought-after modulation technique in a high mobility scenario. Sparse code multiple access (SCMA) is an attractive code-domain non-orthogonal multiple access (NOMA) technique. Recently a code-domain NOMA approach for OTFS, named OTFS-SCMA, is proposed. OTFS-SCMA is a promising framework that meets the demands of high mobility and massive connectivity. This paper presents a channel estimation technique based on the convolutional sparse coding (CSC) approach for OTFS-SCMA in the uplink. The channel estimation task is formulated as a CSC problem following a careful rearrangement of the OTFS input-output relation. We use an embedded pilot-aided sparse-pilot structure that enjoys the features of both OTFS and SCMA. The existing channel estimation techniques for OTFS in multi-user scenarios for uplink demand extremely high overhead for pilot and guard symbols, proportional to the number of users. The proposed method maintains a minimal overhead equivalent to a single user without compromising on the estimation error. The results show that the proposed channel estimation algorithm is very efficient in bit error rate (BER), normalized mean square error (NMSE), and spectral efficiency (SE).

Unmanned-aerial-vehicles (UAVs) based communications are envisioned to play an important role in 5G and beyond 5G (B5G) systems. UAV-to-ground communications in urban cities are often characterized by highly dynamic propagation environments that can be described by composite fading channels. Most of the UAV-to-ground systems are based on first order (FO) performance evaluation, however the models based on FO statistics are insufficient for characterization of time variant fading channels. We provide comprehensive mathematical framework for the second order (SO) statistics over double-scattered, double-shadowed (DS-DS) fading channels, modeled as the product of double Nakagami-m (DN) and double inverse Gamma (DIG) random processes (RPs). In particular, we obtained exact mathematical expressions for average fade duration (AFD) and level crossing rate (LCR) of the proposed UAV-to-ground channel model. Moreover, the exact, integral form SO statistical expressions are approximated by Laplace Integration (LI) and exponential LI in order to provide closed form, easily computing mathematical expressions. Numerical results show that approximate and exact results are fitting well, especially for higher output threshold values. The impact of DS-DS fading severities on the SO statistics are well investigated. Furthermore, the proposed method is extended to analyze SO performances for the selection scenario of UAV with the highest signal level from among N-UAVs links.

Abstract This paper presents a sparse code multiple access (SCMA) system with massive antennas at the base station. This system is referred to as M‐SCMA system. A spectrally‐efficient and massive access next‐generation wireless network is realized through massive antennas and non‐orthogonal SCMA techniques. Two detection algorithms, namely, modified message passing algorithm (MMPA) and extended message passing algorithm (EMPA) are proposed to detect multiple users' symbols in M‐SCMA. A deep learning (DL)‐based detection scheme is also proposed for M‐SCMA so as to avoid channel estimation and to lower the detection complexity. Numerical results show that the DL‐based detection has similar performance as MMPA even when the channel information is not estimated explicitly. Furthermore, authors also establish the sum rate trade‐off between SCMA and orthogonal multiple access in a massive antenna system. The impact of various M‐SCMA parameters such as the number of antennas and the overloading factor, on the proposed DL, MMPA, and EMPA‐based detection are also investigated.

Orthogonal time frequency space (OTFS) modulation is a two-dimensional (2-D) modulation technique that has the potential to overcome the challenges faced by orthogonal frequency division multiplexing (OFDM) in high Doppler environments. The performance of OTFS in a multi-user scenario with orthogonal multiple access (OMA) techniques has been impressive. Due to the requirement of massive connectivity in 5G and beyond, it is essential to devise and examine the OTFS system with the existing non-orthogonal multiple access (NOMA) techniques. This paper proposes a multi-user OTFS system based on a code-domain NOMA technique called sparse code multiple access (SCMA). This system is referred to as the OTFS-SCMA model. The framework for OTFS-SCMA is designed for both downlink and uplink. First, the sparse SCMA codewords are strategically placed on the delay-Doppler plane. The overall overloading factor of the OTFS-SCMA system is equal to that of the underlying basic SCMA system. The receiver in downlink performs the detection in two sequential phases: first, the conventional OTFS detection using the method of linear minimum mean square error (LMMSE) estimation, and then the SCMA detection. We propose a single-phase detector based on a message-passing algorithm (MPA) to detect multiple users’ symbols for the uplink. The expressions for the asymptotic diversity orders of the proposed OTFS-SCMA system are derived for downlink and uplink. OTFS-SCMA provides a significant diversity gain over other multiple access systems for OTFS. Based on the diversity analysis, an algorithm is proposed to devise an optimal codeword allocation scheme. The performance of the proposed OTFS-SCMA system is validated through extensive simulations both in downlink and uplink. We consider delay-Doppler planes of different parameters and various SCMA systems of overloading factor up to 200%. The performance of OTFS-SCMA is compared with those of the existing OTFS-OMA, OFDM-SCMA and OTFS-power-domain (PD)-NOMA techniques. The analysis of OTFS-SCMA with channel estimation is also presented along with the BER performance.

Terahertz (THz) band communication is considered as a promising technology for 6G wireless technology due to availability of large bandwidth at THz bands. Most receivers for symbol detection need channel state information (CSI), however estimating CSI is complex for THz bands. Hence, in this paper, we design and analyze low complexity THz receiver without considering the CSI for symbol detection. The rate adaptation for THz communication systems is also analysed for the distance attenuation. Since, at THz the performance is sensitive to distance. We consider modified transmitted-reference (MTR) pulsed and optimal lth-order energy detector (OED) based THz communication system. Simulations are carried out to show effectiveness of MTR and OED for pulse-based THz band wireless systems by considering the effect of distance, data rate, and absorption by molecular gases.

Diffusion based molecular communication (DBMC) acts as an indispensable element of nano-networking and Internet of things based applications. In DBMC, bio-molecules serve as information carriers for nanoscale systems. However, performance of the DBMC is limited by the following factors; nonlinear channel impulse response (CIR) of the DBMC which makes the mapping between the received symbols and transmitted symbols nonlinear, inter-symbol-interference (ISI) due to long tail of CIR of the DBMC which degrades the overall bit error rate (BER) performance, and stochastic movement of the molecules which introduces noise at the receiver. Most of the prior works on receiver design for molecular communication (MC) are based on linear detection algorithms, which are sub-optimal for the nonlinear DBMC system. Thus, to improve the performance of MC system, for the first time, we propose adaptive nonlinear receiver for MC based on the kernel least mean square algorithm, in the reproducing kernel Hilbert space. Numerical results show that the proposed nonlinear receiver improves BER performance as compared to the existing linear detection algorithms-based receiver.

In impulse radio ultra-wideband (IR-UWB) literature, simple transmitted reference (TR) autocorrelation receiver has been analyzed for data symbol detection. However, TR receiver is neither energy efficient nor data transmission rate efficient due to transmission of two pulses per data symbol. In this paper, we propose an iterative TR (ITR) UWB receiver for the joint time of arrival (ToA) and data symbol estimation. The proposed ITR receiver uses only single reference pulse for a burst of data symbols. The ITR receiver estimates a new reference pulse from the burst of data symbols' pulses to further enhance the signal-to-noise ratio (SNR) of UWB system. Hence, the proposed ITR receiver has high energy efficiency and data rate transmission efficiency with improved SNR as compared to a TR receiver. Further, the proposed ITR receiver is implemented at sub-Nyquist rate to overcome the high sampling rate analog-to- digital converter. The ITR receiver's ToA and data symbol estimation performance is analyzed for the binary phase shift keying modulated UWB system in standard IEEE 802.15.4a channels in the presence of additive white Gaussian noise.

Ultra-wide band (UWB) ranging requires precise estimation of time-of-arrival (ToA) of first path (FP) signal. In UWB communication, estimation of ToA is challenging in the presence of impulsive interference and multipath. Generally in literature, ToA is estimated using threshold crossing techniques. Although threshold based ToA methods are simple, optimal threshold value is hard to determine in practice. In this paper, we propose joint estimation of ToA and data symbols by exploiting the cluster-sparsity of the received UWB signal in the presence of impulsive interference. The proposed receiver structure enhances the signal-to- noise ratio at the receiver output by mitigating the impulsive interference and barring inter-clusters noise accumulation. The proposed method is also free from any threshold, training, and optimization process. Robustness of the proposed algorithm is validated for standardized IEEE 802.15.4a channel models in the presence of both additive white Gaussian noise and impulsive interference.

Ultra-wide band (UWB) communication is a viable solution for Wireless Sensor Network (WSN) and Internet of Things (IoT) due to low cost and low power requirement. However, UWB transceiver design is more complex due to large bandwidth and precise synchronization requirement. In this paper, we propose a simple peak detection based non-coherent UWB receiver, suitable for low data rate WSN and IoT based applications. The proposed receiver divides each data symbol frame duration into smaller multiple time windows. In each time window, peak of received signal is detected independently using threshold comparison. The transmitted signal is detected in a frame by employing decisions on all multiple time windows. From simulation, it is observed that the proposed receiver outperforms existing non-coherent receivers. The performance analysis of the proposed receiver is carried out by using time hopping pulse position modulation (TH-PPM) UWB signal in additive white Gaussian noise (AWGN), multipath communication using the existing IEEE 802.15.4a standard.