Search for: All records

Massive MIMO systems can achieve high spectrum and energy efficiency in downlink (DL) based on accurate estimate of channel state information (CSI). Existing works have developed learning-based DL CSI estimation that lowers uplink feedback overhead. One often overlooked problem is the limited number of DL pilots available for CSI estimation. One proposed solution leverages temporal CSI coherence by utilizing past CSI estimates and only sending CSI-reference symbols (CSIRS) for partial arrays to preserve CSI recovery performance. Exploiting CSI correlations, FDD channel reciprocity is helpful to base stations with direct access to uplink CSI. In this work, we propose a new learning-based feedback architecture and a reconfigurable CSI-RS placement scheme to reduce DL CSI training overhead and to improve encoding efficiency of CSI feedback. Our results demonstrate superior performance in both indoor and outdoor scenarios by the proposed framework for CSI recovery at substantial reduction of computation power and storage requirements at UEs. 

Multiuser MIMO (MU-MIMO) technologies can help provide rapidly growing needs for high data rates in modern wireless networks. Co-channel interference (CCI) among users in the same resource-sharing group (RSG) presents a serious user scheduling challenge to achieve high overall MU-MIMO capacity. Since CCI is closely related to correlation among spatial user channels, it would be natural to schedule co-channel user groups with low inter-user channel correlation. Yet, establishing RSGs with low co-channel correlations for large user populations is an NP-hard problem. More practically, user scheduling for wideband channels exhibiting distinct channel characteristics in each frequency band remains an open question. In this work, we proposed a novel wideband user grouping and scheduling algorithm named SC-MS. The proposed SC-MS algorithm first leverages spectral clustering to obtain a preliminary set of user groups. Next, we apply a post-processing step to identify user cliques from the preliminary groups to further mitigate CCI. Our last step groups users into RSGs for scheduling such that the sum of user clique sizes across the multiple frequency bands is maximized. Simulation results demonstrate network performance gain over benchmark methods in terms of sum rate and fairness. 

Wireless links using massive MIMO transceivers are vital for next generation wireless communications networks. Precoding in Massive MIMO transmission requires accurate downlink channel state information (CSI). Many recent works have effectively applied deep learning (DL) to jointly train UE-side compression networks for delay domain CSI and a BS-side decoding scheme. Vitally, these works assume that the full delay domain CSI is available at the UE, but in reality, the UE must estimate the delay domain based on a limited number of frequency domain pilots. In this work, we propose a linear pilot-to-delay estimator (P2DE) that acquires the truncated delay CSI via sparse frequency pilots. We show the accuracy of the P2DE under frequency downsampling, and we demonstrate the P2DE’s efficacy when utilized with existing CSI estimation networks. Additionally, we propose to use trainable compressed sensing (CS) networks in a differential encoding network for time-varying CSI estimation, and we propose a new network, MarkovNet-ISTA-ENet (MN-IE), which combines a CS network for initial CSI estimation and multiple autoencoders to estimate the error terms. We demonstrate that MN-IE has better asymptotic performance than networks comprised of only one type of network. 

Abstract Hyperspectral imaging has broad applications and impacts in areas including environmental science, weather, and geo/space exploration. The intrinsic spectral–spatial structures and potential multi-level features in different frequency bands make multilayer graph an intuitive model for hyperspectral images (HSI). To study the underlying characteristics of HSI and to take the advantage of graph signal processing (GSP) tools, this work proposes a multilayer graph spectral analysis for hyperspectral images based on multilayer graph signal processing (M-GSP). More specifically, we present multilayer graph (MLG) models and tensor representations for HSI. By exploring multilayer graph spectral space, we develop MLG-based methods for HSI applications, including unsupervised segmentation and supervised classification. Our experimental results demonstrate the strength of M-GSP in HSI processing and spectral–spatial information extraction. 

The past decade has witnessed the rising dominance of deep learning and artificial intelligence in a wide range of applications. In particular, the ocean of wireless smartphones and IoT devices continue to fuel the tremendous growth of edge/cloudbased machine learning (ML) systems including image/speech recognition and classification. To overcome the infrastructural barrier of limited network bandwidth in cloud ML, existing solutions have mainly relied on traditional compression codecs such as JPEG that were historically engineered for humanend users instead of ML algorithms. Traditional codecs do not necessarily preserve features important to ML algorithms under limited bandwidth, leading to potentially inferior performance. This work investigates application-driven optimization of programmable commercial codec settings for networked learning tasks such as image classification. Based on the foundation of variational autoencoders (VAEs), we develop an end-to-end networked learning framework by jointly optimizing the codec and classifier without reconstructing images for given data rate (bandwidth). Compared with standard JPEG codec, the proposed VAE joint compression and classification framework achieves classification accuracy improvement by over 10% and 4%, respectively, for CIFAR-10 and ImageNet-1k data sets at data rate of 0.8 bpp. Our proposed VAE-based models show 65%􀀀99% reductions in encoder size,  1.5􀀀 13.1 improvements in inference speed and 25%􀀀99% savings in power compared to baseline models. We further show that a simple decoder can reconstruct images with sufficient quality without compromising classification accuracy.