On removing potential redundant constraints for SVOR learning

Joint Metric Learning-Based Class-Specific Representation for Image Set Classification

With the rapid advances in digital imaging and communication technologies, recently image set classification has attracted significant attention and has been widely used in many real-world scenarios. As an effective technology, the class-specific representation theory-based methods have demonstrated their superior performances. However, this type of methods either only uses one gallery set to measure the gallery-to-probe set distance or ignores the inner connection between different metrics, leading to the learned distance metric lacking robustness, and is sensitive to the size of image sets. In this article, we propose a novel joint metric learning-based class-specific representation framework (JMLC), which can jointly learn the related and unrelated metrics. By iteratively modeling probe set and related or unrelated gallery sets as affine hull, we reconstruct this hull sparsely or collaboratively over another image set. With the obtained representation coefficients, the combined metric between the query set and the gallery set can then be calculated. In addition, we also derive the kernel extension of JMLC and propose two new unrelated set constituting strategies. Specifically, kernelized JMLC (KJMLC) embeds the gallery sets and probe sets into the high-dimensional Hilbert space, and in the kernel space, the data become approximately linear separable. Extensive experiments on seven benchmark databases show the superiority of the proposed methods to the state-of-the-art image set classifiers.

Wearable Device-Based Intelligent Patrol Inspection System Design and Implementation

The traditional on-site operation of power stations includes inspection and maintenance. However, it heavily relies on experience for maintenance. Most on-site operation and maintenance data are text records. On the one hand, the data processing is tedious for experience to affect the safe on-site operation. On the other hand, we usually cannot give full consideration to the value of maintenance experience, so that the corresponding efficiency is very low. Therefore, this paper proposes a wearable device based remote and intelligent patrol inspection system that uses the cloud video transmission mode of both public and private clouds to realize the video connection between the power stations and the remote diagnosis center and uses the wearable devices for real experience. In this way, the authors can simulate real operation guidance and safety supervision, etc. so as to realize the remote management patrol operations, improve the fault detection efficiency, and improve equipment reliability.

Does the built environment of settlements affect our sentiments? A multi-level and non-linear analysis of Xiamen, China, using social media data

Introduction

Humans spend most of their time in settlements, and the built environment of settlements may affect the residents' sentiments. Research in this field is interdisciplinary, integrating urban planning and public health. However, it has been limited by the difficulty of quantifying subjective sentiments and the small sample size.

Methods

This study uses 147,613 Weibo text check-ins in Xiamen from 2017 to quantify residents' sentiments in 1,096 neighborhoods in the city. A multilevel regression model and gradient boosting decision tree (GBDT) model are used to investigate the multilevel and nonlinear effects of the built environment of neighborhoods and subdistricts on residents' sentiments.

Results

The results show the following: (1) The multilevel regression model indicates that at the neighborhood level, a high land value, low plot ratio, low population density, and neighborhoods close to water are more likely to improve the residents' sentiments. At the subdistrict level, more green space and commercial land, less industry, higher building density and road density, and a smaller migrant population are more likely to promote positive sentiments. Approximately 19% of the total variance in the sentiments occurred among subdistricts. (2) The proportion of green space and commercial land, and the density of buildings and roads are linearly correlated with residents' sentiments. The land value is a basic need and exhibits a nonlinear correlation with sentiments. The plot ratio, population density, and the proportions of industrial land and the migrant population are advanced needs and are nonlinearly correlated with sentiments.

Discussion

The quantitative analysis of sentiments enables setting a threshold of the influence of the built environment on residents' sentiments in neighborhoods and surrounding areas. Our results provide data support for urban planning and implementing targeted measures to improve the living environment of residents.

Edge-Enabled Heart Rate Estimation from Multisensor PPG Signals

Heart rate (HR) estimation from multisensor PPG signals suffers from the dilemma of inconsistent computation results, due to the prevalence of bio-artifacts (BAs). Furthermore, advancements in edge computing have shown promising results from capturing and processing diversified types of sensing signals using the devices of Internet of Medical Things (IoMT). In this paper, an edge-enabled method is proposed to estimate HRs accurately and with low latency from multisensor PPG signals captured by bilateral IoMT devices. First, we design a real-world edge network with several resource-constrained devices, divided into collection edge nodes and computing edge nodes. Second, a self-iteration RR interval calculation method, at the collection edge nodes, is proposed leveraging the inherent frequency spectrum feature of PPG signals and preliminarily eliminating the influence of BAs on HR estimation. Meanwhile, this part also reduces the volume of sent data from IoMT devices to compute edge nodes. Afterward, at the computing edge nodes, a heart rate pool with an unsupervised abnormal detection method is proposed to estimate the average HR. Experimental results show that the proposed method outperforms traditional approaches which rely on a single PPG signal, attaining better results in terms of the consistency and accuracy for HR estimation. Furthermore, at the designed edge network, our proposed method processes a 30 s PPG signal to obtain an HR, consuming only 4.24 s of computation time. Hence, the proposed method is of significant value for the low-latency applications in the field of IoMT healthcare and fitness management.

Medical image registration utilizing tissue P systems

The tissue P system (TPS) possesses intrinsic attributes of parallel execution in comprehensive data and instruction space, which provides fast convergence during the transition from local to global optima. Method- In this study, we have proposed and built a TPSysIR framework using the TPS for image registration that optimizes upon the mutual information (MI) similarity metric to find a global solution. Result- The model was tested on single- and multimodal brain MRI scans and other prominent optimization-based image registration techniques. Conclusion- Results show that, among all methods, TPSysIR provides better MI values with minimum deviation in a range of experiment setups conducted iteratively.

Heart Rate Estimation in Sports Based on Multi-Sensor Data for Sports Intensity Prediction

The heart rate (HR) is the most common measurement of the cardiovascular system. It reflects not only the cardiovascular function, but also the degree of recovery, and has high reliability. The heart rate monitoring can be used in athlete selection, sports training, medical supervision, and fitness to avoid the blindness of exercise intensity arrangement, provide an objective quantitative standard for scientific fitness, and improve the sports performance through monitoring sports intensity. In order to accurately predict the sports intensity, this paper adopts ECG signals and pulse wave to learn an ordinal regression model that can utilize the order relation between different sports intensity level. The experimental results have demonstrated the effectiveness of the proposed sports intensity method.

Manifold learning for novelty detection and its application in gesture recognition

As a state-of-the-art novelty detection method, Kernel Null Foley–Sammon Transform (KNFST) could identify multiple known classes and detect novelties from an unknown class via a single model. However, KNFST only captures the global information of the training set. The local geometrical structure is neglected. In this paper, a manifold is incorporated into KNFST to solve this issue. First, we use manifold graphs to depict the local structure for within-class scatter and total scatter. Second, the training samples from the same class are mapped into a single point in null space via null projected directions (NPDs). The proposed method can overcome the weakness of KNFST caused by ignoring local geometrical structure in the class. The experimental results on several toy and benchmark datasets show that manifold learning novelty detection (MLND) is superior to KNFST.