Direct time-of-flight (dToF) sensors are promising for next-generation on-device 3D sensing. However, to achieve the sufficient signal-to-noise-ratio (SNR) in a compact module, the dToF data has limited spatial resolution (e.g., ~20x30 for iPhone dToF), and it requires a super-resolution step before being passed to downstream tasks. In this paper, we solve this super-resolution problem by fusing the low-resolution dToF data with the corresponding high-resolution RGB guidance. Unlike the conventional RGB-guided depth enhancement approaches which perform the fusion in a per-frame manner, we propose the first multi-frame fusion scheme to mitigate the spatial ambiguity resulting from the low-resolution dToF imaging. In addition, dToF sensors provide unique depth histogram information for each local patch, and we incorporate this dToF-specific feature in our network design to further alleviate spatial ambiguity. To evaluate our models on complex dynamic indoor environments and to provide a large-scale dToF sensor dataset, we introduce DyDToF, the first synthetic RGB-dToF video dataset that features dynamic objects and a realistic dToF simulator following the physical imaging process. We believe the methods and dataset are beneficial to a broad community as dToF depth sensing is becoming mainstream on mobile devices.
The current paper implements a methodology for automatically detecting vehicle maneuvers from vehicle telemetry data under naturalistic driving settings. Previous approaches have treated vehicle maneuver detection as a classification problem, although both time series segmentation and classification are required since input telemetry data is continuous. Our objective is to develop an end-to-end pipeline for frame-by-frame annotation of naturalistic driving studies videos into various driving events including stop and lane keeping events, lane changes, left-right turning movements, and horizontal curve maneuvers. To address the time series segmentation problem, the study developed an Energy Maximization Algorithm (EMA) capable of extracting driving events of varying durations and frequencies from continuous signal data. To reduce overfitting and false alarm rates, heuristic algorithms were used to classify events with highly variable patterns such as stops and lane-keeping. To classify segmented driving events, four machine learning models were implemented, and their accuracy and transferability were assessed over multiple data sources. The duration of events extracted by EMA were comparable to actual events, with accuracies ranging from 59.30% (left lane change) to 85.60% (lane-keeping). Additionally, the overall accuracy of the 1D-convolutional neural network model was 98.99%, followed by the Long-short-term-memory model at 97.75%, then random forest model at 97.71%, and the support vector machine model at 97.65%. These model accuracies where consistent across different data sources. The study concludes that implementing a segmentation-classification pipeline significantly improves both the accuracy for driver maneuver detection and transferability of shallow and deep ML models across diverse datasets.
Background subtraction is a fundamental task in computer vision with numerous real-world applications, ranging from object tracking to video surveillance. Dynamic backgrounds poses a significant challenge here. Supervised deep learning-based techniques are currently considered state-of-the-art for this task. However, these methods require pixel-wise ground-truth labels, which can be time-consuming and expensive. In this work, we propose a weakly supervised framework that can perform background subtraction without requiring per-pixel ground-truth labels. Our framework is trained on a moving object-free sequence of images and comprises two networks. The first network is an autoencoder that generates background images and prepares dynamic background images for training the second network. The dynamic background images are obtained by thresholding the background-subtracted images. The second network is a U-Net that uses the same object-free video for training and the dynamic background images as pixel-wise ground-truth labels. During the test phase, the input images are processed by the autoencoder and U-Net, which generate background and dynamic background images, respectively. The dynamic background image helps remove dynamic motion from the background-subtracted image, enabling us to obtain a foreground image that is free of dynamic artifacts. To demonstrate the effectiveness of our method, we conducted experiments on selected categories of the CDnet 2014 dataset and the I2R dataset. Our method outperformed all top-ranked unsupervised methods. We also achieved better results than one of the two existing weakly supervised methods, and our performance was similar to the other. Our proposed method is online, real-time, efficient, and requires minimal frame-level annotation, making it suitable for a wide range of real-world applications.
This paper considers intelligent reflecting surface (IRS)-aided simultaneous wireless information and power transfer (SWIPT) in a multi-user multiple-input single-output (MISO) interference channel (IFC), where multiple transmitters (Txs) serve their corresponding receivers (Rxs) in a shared spectrum with the aid of IRSs. Our goal is to maximize the sum rate of the Rxs by jointly optimizing the transmit covariance matrices at the Txs, the phase shifts at the IRSs, and the resource allocation subject to the individual energy harvesting (EH) constraints at the Rxs. Towards this goal and based on the well-known power splitting (PS) and time switching (TS) receiver structures, we consider three practical transmission schemes, namely the IRS-aided hybrid TS-PS scheme, the IRS-aided time-division multiple access (TDMA) scheme, and the IRS-aided TDMA-D scheme. The latter two schemes differ in whether the Txs employ deterministic energy signals known to all the Rxs. Despite the non-convexity of the three optimization problems corresponding to the three transmission schemes, we develop computationally efficient algorithms to address them suboptimally, respectively, by capitalizing on the techniques of alternating optimization (AO) and successive convex approximation (SCA). Moreover, we conceive feasibility checking methods for these problems, based on which the initial points for the proposed algorithms are constructed. Simulation results demonstrate that our proposed IRS-aided schemes significantly outperform their counterparts without IRSs in terms of sum rate and maximum EH requirements that can be satisfied under various setups. In addition, the IRS-aided hybrid TS-PS scheme generally achieves the best sum rate performance among the three proposed IRS-aided schemes, and if not, increasing the number of IRS elements can always accomplish it.
Active learning is perhaps most naturally posed as an online learning problem. However, prior active learning approaches with deep neural networks assume offline access to the entire dataset ahead of time. This paper proposes VeSSAL, a new algorithm for batch active learning with deep neural networks in streaming settings, which samples groups of points to query for labels at the moment they are encountered. Our approach trades off between uncertainty and diversity of queried samples to match a desired query rate without requiring any hand-tuned hyperparameters. Altogether, we expand the applicability of deep neural networks to realistic active learning scenarios, such as applications relevant to HCI and large, fractured datasets.
Functional electrical stimulation (FES) has been increasingly integrated with other rehabilitation devices, including robots. FES cycling is one of the common FES applications in rehabilitation, which is performed by stimulating leg muscles in a certain pattern. The appropriate pattern varies across individuals and requires manual tuning which can be time-consuming and challenging for the individual user. Here, we present an AI-based method for finding the patterns, which requires no extra hardware or sensors. Our method has two phases, starting with finding model-based patterns using reinforcement learning and detailed musculoskeletal models. The models, built using open-source software, can be customised through our automated script and can be therefore used by non-technical individuals without extra cost. Next, our method fine-tunes the pattern using real cycling data. We test our both in simulation and experimentally on a stationary tricycle. In the simulation test, our method can robustly deliver model-based patterns for different cycling configurations. The experimental evaluation shows that our method can find a model-based pattern that induces higher cycling speed than an EMG-based pattern. By using just 100 seconds of cycling data, our method can deliver a fine-tuned pattern that gives better cycling performance. Beyond FES cycling, this work is a showcase, displaying the feasibility and potential of human-in-the-loop AI in real-world rehabilitation.
The automatic extraction of information from Cyber Threat Intelligence (CTI) reports is crucial in risk management. The increased frequency of the publications of these reports has led researchers to develop new systems for automatically recovering different types of entities and relations from textual data. Most state-of-the-art models leverage Natural Language Processing (NLP) techniques, which perform greatly in extracting a few types of entities at a time but cannot detect heterogeneous data or their relations. Furthermore, several paradigms, such as STIX, have become de facto standards in the CTI community and dictate a formal categorization of different entities and relations to enable organizations to share data consistently. This paper presents STIXnet, the first solution for the automated extraction of all STIX entities and relationships in CTI reports. Through the use of NLP techniques and an interactive Knowledge Base (KB) of entities, our approach obtains F1 scores comparable to state-of-the-art models for entity extraction (0.916) and relation extraction (0.724) while considering significantly more types of entities and relations. Moreover, STIXnet constitutes a modular and extensible framework that manages and coordinates different modules to merge their contributions uniquely and exhaustively. With our approach, researchers and organizations can extend their Information Extraction (IE) capabilities by integrating the efforts of several techniques without needing to develop new tools from scratch.
Despite rapid advancement in recent years, current speech enhancement models often produce speech that differs in perceptual quality from real clean speech. We propose a learning objective that formalizes differences in perceptual quality, by using domain knowledge of acoustic-phonetics. We identify temporal acoustic parameters -- such as spectral tilt, spectral flux, shimmer, etc. -- that are non-differentiable, and we develop a neural network estimator that can accurately predict their time-series values across an utterance. We also model phoneme-specific weights for each feature, as the acoustic parameters are known to show different behavior in different phonemes. We can add this criterion as an auxiliary loss to any model that produces speech, to optimize speech outputs to match the values of clean speech in these features. Experimentally we show that it improves speech enhancement workflows in both time-domain and time-frequency domain, as measured by standard evaluation metrics. We also provide an analysis of phoneme-dependent improvement on acoustic parameters, demonstrating the additional interpretability that our method provides. This analysis can suggest which features are currently the bottleneck for improvement.
Computational predictions of mass spectra from molecules have enabled the discovery of clinically relevant metabolites. However, such predictive tools are still limited as they occupy one of two extremes, either operating (a) by fragmenting molecules combinatorially with overly rigid constraints on potential rearrangements and poor time complexity or (b) by decoding lossy and nonphysical discretized spectra vectors. In this work, we introduce a new intermediate strategy for predicting mass spectra from molecules by treating mass spectra as sets of chemical formulae, which are themselves multisets of atoms. After first encoding an input molecular graph, we decode a set of chemical subformulae, each of which specify a predicted peak in the mass spectra, the intensities of which are predicted by a second model. Our key insight is to overcome the combinatorial possibilities for chemical subformulae by decoding the formula set using a prefix tree structure, atom-type by atom-type, representing a general method for ordered multiset decoding. We show promising empirical results on mass spectra prediction tasks.
A recent development in machine learning - physics-informed deep learning (PIDL) - presents unique advantages in transportation applications such as traffic state estimation. Consolidating the benefits of deep learning (DL) and the governing physical equations, it shows the potential to complement traditional sensing methods in obtaining traffic states. In this paper, we first explain the conservation law from the traffic flow theory as ``physics'', then present the architecture of a PIDL neural network and demonstrate its effectiveness in learning traffic conditions of unobserved areas. In addition, we also exhibit the data collection scenario using fog computing infrastructure. A case study on estimating the vehicle velocity is presented and the result shows that PIDL surpasses the performance of a regular DL neural network with the same learning architecture, in terms of convergence time and reconstruction accuracy. The encouraging results showcase the broad potential of PIDL for real-time applications in transportation with a low amount of training data.