Subject: Cutting-Edge Advancements in Wireless Communication, Sensing, and Signal Processing

Hi Elman,

This newsletter covers recent preprints exploring cutting-edge advancements in wireless communication, sensing, and signal processing, with a focus on novel hardware and algorithms. We'll delve into breakthroughs in channel estimation, signal recovery, hardware designs, and specific applications like ISAC, healthcare, and beyond.

General Overview

This collection of preprints showcases significant progress in wireless communication, sensing, and signal processing. Several papers concentrate on enhancing channel estimation and signal recovery in challenging environments. Liang et al. (2024) (Liang et al., 2024) introduce a direct multipath-based SLAM method, bypassing traditional channel estimation by directly using received radio signals for improved localization and mapping. Del Hougne (2024) (del Hougne, 2024) presents a physics-compliant diagonal representation for channels parameterized by beyond-diagonal reconfigurable intelligent surfaces (BD-RIS), simplifying channel estimation and optimization for these advanced architectures. Hamadouche and Sellathurai (2024) (Hamadouche & Sellathurai, 2024) propose an efficient dual-blind deconvolution framework using ADMM for joint radar-communication systems, improving channel estimation and signal recovery, particularly in 5G mmWave networks. Magalhães and de Almeida (2024) (Magalhães & de Almeida, 2024) develop semi-blind receivers for hybrid reflecting and sensing RIS, enabling joint symbol and channel estimation.

Novel hardware and system designs also feature prominently. Gao et al. (2024) (Gao et al., 2024) investigate movable antenna (MA) technology in wireless-powered NOMA networks, showing substantial throughput gains. Xiao et al. (2024) (Xiao et al., 2024) explore channel estimation for MA-aided wideband systems. Zhou et al. (2024) (Zhou et al., 2024) examine the use of fluid antennas in near-field ISAC systems. Khosroshahi et al. (2024) (Khosroshahi et al., 2024) propose superimposing PRS and PDSCH signals for ISAC systems, enhancing spectral efficiency. Yu et al. (2024) (Yu et al., 2024) introduce cooperative passive sensing for mmWave blockage prediction.

Specific applications and algorithmic advancements are also addressed. Yang et al. (2024) (Yang et al., 2024) present an integrated sensing, computing, and semantic communication framework for e-healthcare. R R et al. (2024) (R R et al., 2024) analyze caching and handovers in 3D UAV networks. Bakhit et al. (2024) (Bakhit et al., 2024) investigate estimation errors in MIMO systems. Kloob et al. (2024) (Kloob et al., 2024) propose KLD-based waveform design for multi-user multi-target ISAC. Oh et al. (2024) (Oh et al., 2024) utilize deep neural networks for beamforming alignment in polarization reconfigurable MISO systems.

The collection extends beyond wireless communication, encompassing signal processing and machine learning. Sheng et al. (2024) (Sheng et al., 2024) compare Voronoi constellations and probabilistic shaping. Gowda et al. (2024) (Gowda et al., 2024) present an sEMG dataset for gesture decoding. Rappaport et al. (2024) (Rappaport et al., 2024) propose a standardized method for presenting radio propagation channel statistics. Guo et al. (2024) (Guo et al., 2024) review beamspace modulation for near-field MIMO.

Finally, innovative sensing applications are showcased, including connected cars for cloud dynamics (Veihelmann et al., 2024), a 6G AI-enabled air interface (Zhang et al., 2024), sleep apnea diagnosis (Wang et al., 2024; Wang et al., 2024), loneliness forecasting (Yang et al., 2024), and nanodiamond-based quantum receivers (Zeng et al., 2024).

Paper Highlights

Big T-Rex: FDR-Controlled Variable Selection for Millions of Variables

Solving FDR-Controlled Sparse Regression Problems with Five Million Variables on a Laptop by Fabian Scheidt, Jasin Machkour, Michael Muma https://arxiv.org/abs/2409.19088

Caption: This graph compares the average RAM allocation of the original T-Rex algorithm with the new Big T-Rex and its two dummy permutation strategies (S1 and S2) as the number of variables increases. Big T-Rex significantly reduces RAM usage, enabling analysis of millions of variables on a standard laptop. The dummy permutation strategies further improve efficiency, with S1 offering the best performance in terms of both RAM usage and computation time.

Reproducibility is paramount in high-dimensional variable selection, particularly in fields like genomics dealing with millions of variables. Existing multivariate FDR-controlling methods often require powerful computing clusters. The Terminating-Random Experiments (T-Rex) selector offered a solution, but its high RAM requirements hindered its usability on standard laptops. Big T-Rex addresses this limitation, enabling large-scale analyses on readily available hardware.

Big T-Rex utilizes memory mapping, storing data on the SSD and accessing it on demand, thus circumventing RAM limitations. Two novel dummy generation strategies based on permuting a reference matrix further reduce RAM and SSD usage compared to the original T-Rex. These strategies preserve the i.i.d. properties of dummies, which are essential for FDR control. The permutations are applied to row and column indices of a reference dummy matrix X_ref, with elements then copied to the working matrix X^(k):

X^(k) = [X Π^(k)(X_ref)], k > 1

where Π^(k) denotes the permutation function for the kth experiment.

Simulations using a linear Gaussian regression model demonstrate Big T-Rex's drastic RAM reduction, up to 88x compared to the original T-Rex. The dummy permutation strategy (DP Big T-Rex S1) achieves the best computation time, a reduction of up to 6x. Big T-Rex maintains FDR control and comparable True Positive Rate (TPR) to the original T-Rex while ensuring the validity of generated dummies. Remarkably, Big T-Rex solves a problem with 5,000,000 variables in 30 minutes on a laptop, democratizing access to large-scale analyses without requiring expensive high-performance computing.

Quantum Receivers for Ubiquitous 6G Radio Access

Towards ubiquitous radio access using nanodiamond based quantum receivers by Qunsong Zeng, Jiahua Zhang, Madhav Gupta, Zhiqin Chu, Kaibin Huang https://arxiv.org/abs/2409.19273

6G wireless communication necessitates innovative solutions for managing numerous base stations and detecting multi-band signals. This paper proposes leveraging nitrogen-vacancy (NV) centers in fluorescent nanodiamonds (FNDs) to create compact, robust multi-user receivers. Each FND, acting as a nano-antenna, exhibits a unique fluorescence pattern in response to microwaves. By modulating microwave signals carrying information, these patterns enable simultaneous demodulation of multiple users' signals, overcoming limitations of traditional antenna arrays.

The researchers demonstrated simultaneous transmission of two uncoded digitally modulated bit streams. Using frequency modulation, they achieved a low bit error ratio (BER) for image transmission. Amplitude modulation resulted in a BER of zero, and the system supports joint amplitude and frequency modulation. Tunable frequency band communication is achieved by detuning NV center spin resonance frequencies, eliminating the need for separate antennas.

The multiple access scheme utilizes a reference-based approach, comparing received fluorescence images with reference images using a minimum mean squared error (MSE) criterion: (x, y) = arg min||It – R(x, y)||, where (x, y) is the detected bit-pair, It is the received image, and R(x, y) is the reference image. A novel reference-free design leverages FND heterogeneity under magnetic field gradients, eliminating reference bit transmission overhead.

A miniaturized FND-receiver device demonstrates the practicality of this technology. While exhibiting a slightly higher BER than the laboratory setup, performance remains acceptable. Increasing fluorescence spots and laser power improves BER, indicating key optimization parameters. This work highlights the potential of quantum sensing in 6G, offering a path towards ubiquitous radio access.

Analog FFTs for Scalable and Efficient Signal Processing

Analog fast Fourier transforms for scalable and efficient signal processing by T. Patrick Xiao, Ben Feinberg, David K. Richardson, Matthew Cannon, Harsha Medu, Vineet Agrawal, Matthew J. Marinella, Sapan Agarwal, Christopher H. Bennett https://arxiv.org/abs/2409.19071

Edge computing requires efficient signal processing, with the Discrete Fourier Transform (DFT) being central to many applications. Digital processors use the Fast Fourier Transform (FFT) algorithm, reducing DFT complexity from O(N²) to O(Nlog₂N). Analog in-memory computing (IMC) has relied on direct matrix-vector multiplication (MVM) with O(N²) scaling. This paper introduces an analog FFT implementation, achieving O(NlogₖN) scaling, mirroring digital counterparts and unlocking scalability and efficiency.

The analog FFT maps the Cooley-Tukey algorithm onto analog IMC architectures. Large DFTs are decomposed into smaller elementary DFTs (radices) computed as analog MVMs on a resistive memory array. Digital processing handles twiddle factor multiplications and data reshaping between stages. Unlike digital FFTs, analog FFT efficiency peaks when the radix size matches the memory array capacity, minimizing analog-to-digital conversions.

Experimental validation used a SONOS memory array, computing analog FFTs on 1D audio and 2D image signals. The researchers achieved a 65,536-point analog DFT, over 1000x larger than previous demonstrations. The analog FFT showed superior precision and robustness compared to direct MVM. The 1D DFT formula:

Xₖ = Σⁿ⁻¹ xₙe⁻ⁱ²πⁿᵏ/ᴺ

where x is the input, X is the frequency spectrum, and N is the signal length, can be expressed as a matrix-vector multiplication: X = Wₙx, where Wₙ is the DFT matrix.

The analog FFT outperformed digital FFTs on a 7-nm AMD Xilinx Versal chip, demonstrating a 15x to 60x energy advantage across practical DFT sizes. This highlights its potential for power-constrained edge processing. The analog FFT's weak scaling with array size—O(NlogₖN)—allows small memory arrays to handle large FFTs, opening avenues for diverse applications. This work demonstrates the feasibility of analog FFTs and provides key design insights, paving the way for energy-efficient signal processing.

Conclusion

This newsletter highlighted a convergence of innovative hardware and algorithmic approaches driving advancements in wireless communication and sensing. Big T-Rex empowers researchers with accessible high-dimensional data analysis on standard laptops, democratizing this crucial capability. The development of nanodiamond-based quantum receivers promises a paradigm shift in 6G, enabling ubiquitous radio access with enhanced flexibility and performance. Finally, the introduction of analog FFTs unlocks unprecedented scalability and energy efficiency for signal processing in edge devices, paving the way for a new era of powerful and resource-efficient applications. These breakthroughs collectively represent significant progress towards more capable and accessible communication and sensing technologies.