arXiv subjects

Sunwoo Kim

Publications and source records attributed to Sunwoo Kim.

At least 91 records · Page 5Linked to original sources

arXiv2021

Personalized Federated Learning over non-IID Data for Indoor Localization

Localization and tracking of objects using data-driven methods is a popular topic due to the complexity in characterizing the physics of wireless channel propagation models. In these modeling approaches, data needs to be gathered to accurately train models, at the same time that user's privacy is maintained. An appealing scheme to cooperatively achieve these goals is known as Federated Learning (FL). A challenge in FL schemes is the presence of non-independent and identically distributed (non-IID) data, caused by unevenly exploration of different areas. In this paper, we consider the use of recent FL schemes to train a set of personalized models that are then optimally fused through Bayesian rules, which makes it appropriate in the context of indoor localization.

cs.LG↗

arXiv2021

Atomic Norm Minimization-based Low-Overhead Channel Estimation for RIS-aided MIMO Systems

Large beam training overhead has been considered as one of main issues in the channel estimation for reconfigurable intelligent surface (RIS)-aided systems. In this paper, we propose an atomic norm minimization (ANM)-based low-overhead channel estimation for RIS-aided multiple-input-multiple-output (MIMO) systems. When the number of beam training is reduced, some multipath signals may not be received during beam training, and this causes channel estimation failure. To solve this issue, the width of beams created by RIS is widened to capture all multipath signals. Pilot signals received during beam training are compiled into one matrix to define the atomic norm of the channel for RIS-aided MIMO systems. Simulation results show that the proposed algorithm outperforms other channel estimation algorithms.

eess.SP↗

arXiv2021

Location-aware Channel Estimation for RIS-aided mmWave MIMO Systems via Atomic Norm Minimization

In this paper, we propose a location-aware channel estimation based on the atomic norm minimization (ANM) for the reconfigurable intelligent surface (RIS)-aided millimeter-wave multiple-input-multiple-output (MIMO) systems. The beam training overhead at the base station (BS) is reduced by the direct beam steering towards the RIS with the location of the BS and the RIS. The RIS beamwidth adaptation is proposed to reduce the beam training overhead at the RIS, and also it enables accurate channel estimation by ensuring the user equipment receives all the multipath components from the RIS. After the beam training, the cascaded effective channel of the RIS-aided MIMO systems is estimated by ANM. Depending on whether the beam training overhead at the BS or at the RIS is reduced or not, the channel is represented as a linear combination of either 1D atoms, 2D atoms, or 3D atoms, and the ANM is applied to estimate the channel. Simulation results show that the proposed location-aware channel estimation via 2D ANM and 3D ANM achieves superior estimation accuracy to benchmarks.

eess.SP↗

arXiv2021

Dirichlet process approach for radio-based simultaneous localization and mapping

Due to 5G millimeter wave (mmWave), spatial channel parameters are becoming highly resolvable, enabling accurate vehicle localization and mapping. We propose a novel method of radio simultaneous localization and mapping (SLAM) with the Dirichlet process (DP). The DP, which can estimate the number of clusters as well as clustering, is capable of identifying the locations of reflectors by classifying signals when such 5G signals are reflected and received from various objects. We generate birth points using the measurements from 5G mmWave signals received by the vehicle and classify objects by clustering birth points generated over time. Each time we use the DP clustering method, we can map landmarks in the environment in challenging situations where false alarms exist in the measurements and change the cardinality of received signals. Simulation results demonstrate the performance of the proposed scheme. By comparing the results with the SLAM based on the Rao-Blackwellized probability hypothesis density filter, we confirm a slight drop in SLAM performance, but as a result, we validate that it has a significant gain in computational complexity.

eess.SP↗

arXiv2021

Cooperative mmWave PHD-SLAM with Moving Scatterers

Using the multiple-model (MM) probability hypothesis density (PHD) filter, millimeter wave (mmWave) radio simultaneous localization and mapping (SLAM) in vehicular scenarios is susceptible to movements of objects, in particular vehicles driving in parallel with the ego vehicle. We propose and evaluate two countermeasures to track vehicle scatterers (VSs) in mmWave radio MM-PHD-SLAM. First, locally at each vehicle, we generate and treat the VS map PHD in the context of Bayesian recursion, and modify vehicle state correction with the VS map PHD. Second, in the global map fusion process at the base station, we average the VS map PHD and upload it with self-vehicle posterior density, compute fusion weights, and prune the target with low Gaussian weight in the context of arithmetic average-based map fusion. From simulation results, the proposed cooperative mmWave radio MM-PHD-SLAM filter is shown to outperform the previous filter in VS scenarios.

eess.SP↗

arXiv2021

Test-Time Adaptation Toward Personalized Speech Enhancement: Zero-Shot Learning with Knowledge Distillation

In realistic speech enhancement settings for end-user devices, we often encounter only a few speakers and noise types that tend to reoccur in the specific acoustic environment. We propose a novel personalized speech enhancement method to adapt a compact denoising model to the test-time specificity. Our goal in this test-time adaptation is to utilize no clean speech target of the test speaker, thus fulfilling the requirement for zero-shot learning. To complement the lack of clean utterance, we employ the knowledge distillation framework. Instead of the missing clean utterance target, we distill the more advanced denoising results from an overly large teacher model, and use it as the pseudo target to train the small student model. This zero-shot learning procedure circumvents the process of collecting users' clean speech, a process that users are reluctant to comply due to privacy concerns and technical difficulty of recording clean voice. Experiments on various test-time conditions show that the proposed personalization method achieves significant performance gains compared to larger baseline networks trained from a large speaker- and noise-agnostic datasets. In addition, since the compact personalized models can outperform larger general-purpose models, we claim that the proposed method performs model compression with no loss of denoising performance.

eess.AS↗

arXiv2021

Results of the 2020 fastMRI Challenge for Machine Learning MR Image Reconstruction

Accelerating MRI scans is one of the principal outstanding problems in the MRI research community. Towards this goal, we hosted the second fastMRI competition targeted towards reconstructing MR images with subsampled k-space data. We provided participants with data from 7,299 clinical brain scans (de-identified via a HIPAA-compliant procedure by NYU Langone Health), holding back the fully-sampled data from 894 of these scans for challenge evaluation purposes. In contrast to the 2019 challenge, we focused our radiologist evaluations on pathological assessment in brain images. We also debuted a new Transfer track that required participants to submit models evaluated on MRI scanners from outside the training set. We received 19 submissions from eight different groups. Results showed one team scoring best in both SSIM scores and qualitative radiologist evaluations. We also performed analysis on alternative metrics to mitigate the effects of background noise and collected feedback from the participants to inform future challenges. Lastly, we identify common failure modes across the submissions, highlighting areas of need for future research in the MRI reconstruction community.

eess.IV↗

arXiv2021

Personalized Speech Enhancement through Self-Supervised Data Augmentation and Purification

Training personalized speech enhancement models is innately a no-shot learning problem due to privacy constraints and limited access to noise-free speech from the target user. If there is an abundance of unlabeled noisy speech from the test-time user, a personalized speech enhancement model can be trained using self-supervised learning. One straightforward approach to model personalization is to use the target speaker's noisy recordings as pseudo-sources. Then, a pseudo denoising model learns to remove injected training noises and recover the pseudo-sources. However, this approach is volatile as it depends on the quality of the pseudo-sources, which may be too noisy. As a remedy, we propose an improvement to the self-supervised approach through data purification. We first train an SNR predictor model to estimate the frame-by-frame SNR of the pseudo-sources. Then, the predictor's estimates are converted into weights which adjust the frame-by-frame contribution of the pseudo-sources towards training the personalized model. We empirically show that the proposed data purification step improves the usability of the speaker-specific noisy data in the context of personalized speech enhancement. Without relying on any clean speech recordings or speaker embeddings, our approach may be seen as privacy-preserving.

eess.AS↗

arXiv2020

Online Exemplar Fine-Tuning for Image-to-Image Translation

Existing techniques to solve exemplar-based image-to-image translation within deep convolutional neural networks (CNNs) generally require a training phase to optimize the network parameters on domain-specific and task-specific benchmarks, thus having limited applicability and generalization ability. In this paper, we propose a novel framework, for the first time, to solve exemplar-based translation through an online optimization given an input image pair, called online exemplar fine-tuning (OEFT), in which we fine-tune the off-the-shelf and general-purpose networks to the input image pair themselves. We design two sub-networks, namely correspondence fine-tuning and multiple GAN inversion, and optimize these network parameters and latent codes, starting from the pre-trained ones, with well-defined loss functions. Our framework does not require the off-line training phase, which has been the main challenge of existing methods, but the pre-trained networks to enable optimization in online. Experimental results prove that our framework is effective in having a generalization power to unseen image pairs and clearly even outperforms the state-of-the-arts needing the intensive training phase.

cs.CV↗

arXiv2020

Transaction-level Model Simulator for Communication-Limited Accelerators

Rapid design space exploration in early design stage is critical to algorithm-architecture co-design for accelerators. In this work, a pre-RTL cycle-accurate accelerator simulator based on SystemC transaction-level modeling (TLM), AccTLMSim, is proposed for convolutional neural network (CNN) accelerators. The accelerator simulator keeps track of each bus transaction between accelerator and DRAM, taking into account the communication bandwidth. The simulation results are validated against the implementation results on the Xilinx Zynq. Using the proposed simulator, it is shown that the communication bandwidth is severely affected by DRAM latency and bus protocol overhead. In addition, the loop tiling is optimized to maximize the performance under the constraint of on-chip SRAM size. Furthermore, a new performance estimation model is proposed to speed up the design space exploration. Thanks to the proposed simulator and performance estimation model, it is possible to explore a design space of millions of architectural options within a few tens of minutes.

cs.AR↗

arXiv2020

Exploiting Diffuse Multipath in 5G SLAM

5G millimeter wave (mmWave) signals can be used to jointly localize the receiver and map the propagation environment in vehicular networks, which is a typical simultaneous localization and mapping (SLAM) problem. Mapping the environment is challenging, due to measurements comprising both specular and diffuse multipath components, and diffuse multipath is usually considered as a perturbation. We here propose a novel method to utilize all available multipath signals from each landmark for mapping and incorporate this into a Poisson multi-Bernoulli mixture for the 5G SLAM problem. Simulation results demonstrate the efficacy of the proposed scheme.

eess.SP↗

arXiv2020

Experimental Demonstration of Location-aware Beam Alignment

The main focus of beam alignment is to find the optimal beam which yields the largest received signal strength (RSS) with faster speed.In this paper, we demonstrate an efficient beam alignment scheme with our testbed. The algorithm we experiment uses the location information for the computation efficient beam alignment.The testbed transmits and receives the 13.8 GHz signal and steers a beam on both transmitter and receiver with various radio frequency (RF) components. The location information is estimated with the indoor positioning module. The experiment shows that the location-aware algorithm significantly reduces the time consumption for beam alignment than the exhaustive search.

eess.SP↗

arXiv2020

5G mmWave Cooperative Positioning and Mapping using Multi-Model PHD Filter and Map Fusion

5G millimeter wave (mmWave) signals can enable accurate positioning in vehicular networks when the base station and vehicles are equipped with large antenna arrays. However, radio-based positioning suffers from multipath signals generated by different types of objects in the physical environment. Multipath can be turned into a benefit, by building up a radio map (comprising the number of objects, object type, and object state) and using this map to exploit all available signal paths for positioning. We propose a new method for cooperative vehicle positioning and mapping of the radio environment, comprising a multiple-model probability hypothesis density filter and a map fusion routine, which is able to consider different types of objects and different fields of views. Simulation results demonstrate the performance of the proposed method.

eess.SP↗

arXiv2020

Boosted Locality Sensitive Hashing: Discriminative Binary Codes for Source Separation

Speech enhancement tasks have seen significant improvements with the advance of deep learning technology, but with the cost of increased computational complexity. In this study, we propose an adaptive boosting approach to learning locality sensitive hash codes, which represent audio spectra efficiently. We use the learned hash codes for single-channel speech denoising tasks as an alternative to a complex machine learning model, particularly to address the resource-constrained environments. Our adaptive boosting algorithm learns simple logistic regressors as the weak learners. Once trained, their binary classification results transform each spectrum of test noisy speech into a bit string. Simple bitwise operations calculate Hamming distance to find the K-nearest matching frames in the dictionary of training noisy speech spectra, whose associated ideal binary masks are averaged to estimate the denoising mask for that test mixture. Our proposed learning algorithm differs from AdaBoost in the sense that the projections are trained to minimize the distances between the self-similarity matrix of the hash codes and that of the original spectra, rather than the misclassification rate. We evaluate our discriminative hash codes on the TIMIT corpus with various noise types, and show comparative performance to deep learning methods in terms of denoising performance and complexity.

eess.AS↗

arXiv2019

Euclidean Matchings in Ultra-Dense Networks

In order to study the fundamental limits of network densification, we look at the spatial spectral efficiency gain achieved when densely deployed communication devices embedded in the $d$-dimensional Euclidean space are optimally `matched' in near-neighbour pairs. In light of recent success in probabilisitc modelling, we study devices distributed uniformly at random in the unit cube which enter into one-on-one contracts with each another. This is known in statistical physics as an Euclidean `matching'. Communication channels each have their own maximal data capacity given by Shannon's theorem. The length of the shortest matching then corresponds to the maximum one-hop capacity on those points. Interference is then added as a further constraint, which is modelled using shapes as guard regions, such as a disk, diametral disk, or equilateral triangle, matched to points, in a similar light to computational geometry. The disk, for example, produces the Delaunay triangulation, while the diametral disk produces a beta-skeleton. We also discuss deriving the scaling limit of both models using the replica method from the physics of disordered systems.

cond-mat.dis-nn↗

arXiv2019

Nearest Neighbor Search-Based Bitwise Source Separation Using Discriminant Winner-Take-All Hashing

We propose an iteration-free source separation algorithm based on Winner-Take-All (WTA) hash codes, which is a faster, yet accurate alternative to a complex machine learning model for single-channel source separation in a resource-constrained environment. We first generate random permutations with WTA hashing to encode the shape of the multidimensional audio spectrum to a reduced bitstring representation. A nearest neighbor search on the hash codes of an incoming noisy spectrum as the query string results in the closest matches among the hashed mixture spectra. Using the indices of the matching frames, we obtain the corresponding ideal binary mask vectors for denoising. Since both the training data and the search operation are bitwise, the procedure can be done efficiently in hardware implementations. Experimental results show that the WTA hash codes are discriminant and provide an affordable dictionary search mechanism that leads to a competent performance compared to a comprehensive model and oracle masking.

eess.AS↗

arXiv2019

Incremental Binarization On Recurrent Neural Networks For Single-Channel Source Separation

This paper proposes a Bitwise Gated Recurrent Unit (BGRU) network for the single-channel source separation task. Recurrent Neural Networks (RNN) require several sets of weights within its cells, which significantly increases the computational cost compared to the fully-connected networks. To mitigate this increased computation, we focus on the GRU cells and quantize the feedforward procedure with binarized values and bitwise operations. The BGRU network is trained in two stages. The real-valued weights are pretrained and transferred to the bitwise network, which are then incrementally binarized to minimize the potential loss that can occur from a sudden introduction of quantization. As the proposed binarization technique turns only a few randomly chosen parameters into their binary versions, it gives the network training procedure a chance to gently adapt to the partly quantized version of the network. It eventually achieves the full binarization by incrementally increasing the amount of binarization over the iterations. Our experiments show that the proposed BGRU method produces source separation results greater than that of a real-valued fully connected network, with 11-12 dB mean Signal-to-Distortion Ratio (SDR). A fully binarized BGRU still outperforms a Bitwise Neural Network (BNN) by 1-2 dB even with less number of layers.

eess.AS↗

arXiv2019

Adaptive Beamwidth Control for mmWave Beam Tracking

Traditional beam tracking methods have severe performance loss under the high mobility and narrow beam scenario. To alleviate the tracking performance degradation, we propose an adaptive beamwidth control for millimeter wave (mmWave) beam tracking. The particle filter is applied to the beam tracking, and the AoA estimation error is approximated with a posterior density function constructed by the particles. The error approximation leads to the adaptive beamwidth control which is implemented by the partial activation of the antenna array. Simulation results show that the proposed algorithm aids the beam tracking to yield a smaller AoA estimation error under the high mobility environments.

eess.SP↗