A hybrid TCN-GRU model for classifying human activities using smartphone inertial signals.

Recognising human activities using smart devices has led to countless inventions in various domains like healthcare, security, sports, etc. Sensor-based human activity recognition (HAR), especially smartphone-based HAR, has become popular among the research community due to lightweight computation and user privacy protection. Deep learning models are the most preferred solutions in developing smartphone-based HAR as they can automatically capture salient and distinctive features from input signals and classify them into respective activity classes. However, in most cases, the architecture of these models needs to be deep and complex for better classification performance. Furthermore, training these models requires extensive computational resources. Hence, this research proposes a hybrid lightweight model that integrates an enhanced Temporal Convolutional Network (TCN) with Gated Recurrent Unit (GRU) layers for salient spatiotemporal feature extraction without tedious manual feature extraction. Essentially, dilations are incorporated into each convolutional kernel in the TCN-GRU model to extend the kernel's field of view without imposing additional model parameters. Moreover, fewer short filters are applied for each convolutional layer to alleviate excess parameters. Despite reducing computational cost, the proposed model utilises dilations, residual connections, and GRU layers for longer-term time dependency modelling by retaining longer implicit features of the input inertial sequences throughout training to provide sufficient information for future prediction. The performance of the TCN-GRU model is verified on two benchmark smartphone-based HAR databases, i.e., UCI HAR and UniMiB SHAR. The model attains promising accuracy in recognising human activities with 97.25% on UCI HAR and 93.51% on UniMiB SHAR. Since the current study exclusively works on the inertial signals captured by smartphones, future studies will explore the generalisation of the proposed TCN-GRU across diverse datasets, including various sensor types, to ensure its adaptability across different applications.

Highest ocean heat in four centuries places Great Barrier Reef in danger.

Mass coral bleaching on the Great Barrier Reef (GBR) in Australia between 2016 and 2024 was driven by high sea surface temperatures (SST)1. The likelihood of temperature-induced bleaching is a key determinant for the future threat status of the GBR2, but the long-term context of recent temperatures in the region is unclear. Here we show that the January-March Coral Sea heat extremes in 2024, 2017 and 2020 (in order of descending mean SST anomalies) were the warmest in 400 years, exceeding the 95th-percentile uncertainty limit of our reconstructed pre-1900 maximum. The 2016, 2004 and 2022 events were the next warmest, exceeding the 90th-percentile limit. Climate model analysis confirms that human influence on the climate system is responsible for the rapid warming in recent decades. This attribution, together with the recent ocean temperature extremes, post-1900 warming trend and observed mass coral bleaching, shows that the existential threat to the GBR ecosystem from anthropogenic climate change is now realized. Without urgent intervention, the iconic GBR is at risk of experiencing temperatures conducive to near-annual coral bleaching3, with negative consequences for biodiversity and ecosystems services. A continuation on the current trajectory would further threaten the ecological function4 and outstanding universal value5 of one of Earth's greatest natural wonders.

Conceptualization of Cloud-Based Motion Analysis and Navigation for Wearable Robotic Applications.

The behavior of pedestrians in a non-constrained environment is difficult to predict. In wearable robotics, this poses a challenge, since devices like lower-limb exoskeletons and active orthoses need to support different walking activities, including level walking and climbing stairs. While a fixed movement trajectory can be easily supported, switches between these activities are difficult to predict. Moreover, the demand for these devices is expected to rise in the years ahead. In this work, we propose a cloud software system for use in wearable robotics, based on geographical mapping techniques and Human Activity Recognition (HAR). The system aims to give context to the surrounding pedestrians by providing hindsight information. The system was partially implemented and tested. The results indicate a viable concept with great extensibility prospects.

MF-Match: A Semi-Supervised Model for Human Action Recognition.

Human action recognition (HAR) technology based on radar signals has garnered significant attention from both industry and academia due to its exceptional privacy-preserving capabilities, noncontact sensing characteristics, and insensitivity to lighting conditions. However, the scarcity of accurately labeled human radar data poses a significant challenge in meeting the demand for large-scale training datasets required by deep model-based HAR technology, thus substantially impeding technological advancements in this field. To address this issue, a semi-supervised learning algorithm, MF-Match, is proposed in this paper. This algorithm computes pseudo-labels for larger-scale unsupervised radar data, enabling the model to extract embedded human behavioral information and enhance the accuracy of HAR algorithms. Furthermore, the method incorporates contrastive learning principles to improve the quality of model-generated pseudo-labels and mitigate the impact of mislabeled pseudo-labels on recognition performance. Experimental results demonstrate that this method achieves action recognition accuracies of 86.69% and 91.48% on two widely used radar spectrum datasets, respectively, utilizing only 10% labeled data, thereby validating the effectiveness of the proposed approach.

Large Language Models for Wearable Sensor-Based Human Activity Recognition, Health Monitoring, and Behavioral Modeling: A Survey of Early Trends, Datasets, and Challenges.

The proliferation of wearable technology enables the generation of vast amounts of sensor data, offering significant opportunities for advancements in health monitoring, activity recognition, and personalized medicine. However, the complexity and volume of these data present substantial challenges in data modeling and analysis, which have been addressed with approaches spanning time series modeling to deep learning techniques. The latest frontier in this domain is the adoption of large language models (LLMs), such as GPT-4 and Llama, for data analysis, modeling, understanding, and human behavior monitoring through the lens of wearable sensor data. This survey explores the current trends and challenges in applying LLMs for sensor-based human activity recognition and behavior modeling. We discuss the nature of wearable sensor data, the capabilities and limitations of LLMs in modeling them, and their integration with traditional machine learning techniques. We also identify key challenges, including data quality, computational requirements, interpretability, and privacy concerns. By examining case studies and successful applications, we highlight the potential of LLMs in enhancing the analysis and interpretation of wearable sensor data. Finally, we propose future directions for research, emphasizing the need for improved preprocessing techniques, more efficient and scalable models, and interdisciplinary collaboration. This survey aims to provide a comprehensive overview of the intersection between wearable sensor data and LLMs, offering insights into the current state and future prospects of this emerging field.

Global distribution pattern of soil phosphorus-cycling microbes under the influence of human activities.

Human activities have profoundly altered the Earth's phosphorus (P) cycling process and its associated microbial communities, yet their global distribution pattern and response to human influences remain unclear. Here, we estimated the abundances of P-cycling genes from 3321 global soil metagenomic samples and mapped the global distribution of five key P-cycling processes, that is, organic phosphoester hydrolysis, inorganic phosphorus solubilization, two-component system, phosphotransferase system, and transporters. Structural equation modeling and random forest analysis were employed to assess the impact of anthropogenic and environmental factors on the abundance of P-cycling genes. Our findings suggest that although less significant than the climate and soil profile, human-related factors, such as economic activities and population, are important drivers for the variations in P-cycling gene abundance. Notably, the gene abundances were increased parallel to the extent of human intervention, but generally at low and moderate levels of human activities. Furthermore, we identified critical genera, such as Pseudomonas and Lysobacter, which were sensitive to the changes in human activities. This study provides insights into the responses of P-cycling microbes to human activities at a global scale, enhancing our understanding of soil microbial P cycling and underscoring the importance of sustainable human activities in the Earth's biogeochemical cycle.

Land use land cover and land surface temperature changes and their relationship with human modification in Islamabad Capital Territory, Pakistan.

Human activities are altering the existing patterns of Land Use Land Cover (LULC) and Land Surface Temperature (LST) on a global scale. However, long-term trends of LULC and LST are largely unknown in many remote mountain areas such as the Karakorum. . The objective of our study therefore was to evaluate the historical changes in land use and land cover (LULC) in an alpine environment located in Islamabad Capital Territory, Pakistan. We used Landsat satellite pictures (namely Landsat 5 TM and Landsat 8 OLI) from the years 1988, 2002, and 2016 and applied the Maximum Likelihood Classification (MLC) approach to categorize land use classes. Land Surface Temperatures (LST) were calculated using the thermal bands (6, 10, and 11) of Landsat series data. The correlation between the Human Modification Index (HMI) and LULC as well as LST was evaluated by utilizing data from Google Earth Engine (GEE). Over the study period, the urbanized area increased by 9.94%, whilst the agricultural and bare soil areas decreased by 3.81% and 3.94%, respectively. The findings revealed a significant change in the LULC with a decrease of 1.99% in vegetation. The highest LST class exhibited a progressive trend, with an increase from 12.27% to 48.48%. Based on the LST analysis, the built-up area shows the highest temperature, followed by the barren, agricultural, and vegetation categories. Similarly, the HMI for different LST categories indicates that higher LST categories have higher levels of human alteration compared to lower LST categories, with a strong correlation (R-value = 0.61) between HMI and LST. The findings can be utilized to promote sustainable urban management and for biodiversity conservation efforts. The work also has the potential of utilizing it to protect delicate ecosystems from human interference and to formulate strategies and regulations for sustainable urban growth, including aspects of land utilization and zoning, reduction of urban heat stress, and urban infrastructure.

Modeling human activity-related spread of the spotted lanternfly (Lycorma delicatula) in the US.

The spotted lanternfly (Lycorma delicatula) has recently spread from its native range to several other countries and forecasts predict that it may become a global invasive pest. In particular, since its confirmed presence in the United States in 2014 it has established itself as a major invasive pest in the Mid-Atlantic region where it is damaging both naturally occurring and commercially important farmed plants. Quarantine zones have been introduced to contain the infestation, but the spread to new areas continues. At present the pathways and drivers of spread are not well-understood. In particular, several human activity related factors have been proposed to contribute to the spread; however, which features of the current spread can be attributed to these factors remains unclear. Here we collect county level data on infestation status and four specific human activity related factors and use statistical methods to determine whether there is evidence for an association between the factors and infestation. Then we construct a network model based on the factors found to be associated with infestation and use it to simulate local spread. We find that the model reproduces key features of the spread 2014 to 2021. In particular, the growth of the main infestation region and the opening of spread corridors in the westward and southwestern directions is consistent with data and the model accurately forecasts the correct infestation status at the county level in 2021 with 81% accuracy. We then use the model to forecast the spread up to 2025 in a larger region. Given that this model is based on a few human activity related factors that can be targeted, it may prove useful to incorporate it into more elaborate predictive forecasting models and in informing management efforts focused on interstate highway transport and garden centers in the US and potentially for current and future invasions elsewhere globally.

Improving human activity classification based on micro-doppler signatures of FMCW radar with the effect of noise.

Nowadays, classifying human activities is applied in many essential fields, such as healthcare, security monitoring, and search and rescue missions. Radar sensor-based human activity classification is regarded as a superior approach in comparison to other techniques, such as visual perception-based methodologies and wearable gadgets. However, noise usually exists throughout the process of extracting raw radar signals, decreasing the quality and reliability of the extracted features. This paper presents a novel method for removing white Gaussian noise from raw radar signals using a denoising algorithm before classifying human activities using a deep convolutional neural network (DCNN). Specifically, the denoising algorithm is used as a preprocessing step to remove white Gaussian noise from the input raw radar signal. After that, a lightweight Cross-Residual Convolutional Neural Network (CRCNN) with adaptable cross-residual connections is suggested for classification. The analysis results show that the denoising algorithm with a range-bin interval of 3 and a cut-threshold value of 3 achieves the best denoising effect. When the denoising algorithm was applied to the dataset, CRCNN improved the right classification rate by up to 10% compared to the recognition results achieved with the original noise-added dataset. Additionally, a comparison of the CRCNN with the denoising algorithm solution with six cutting-edge DCNNs was conducted. The experimental results reveal that the proposed model greatly outperforms the others.

Ecological roles and importance of sharks in the Anthropocene Ocean.

In ecosystems, sharks can be predators, competitors, facilitators, nutrient transporters, and food. However, overfishing and other threats have greatly reduced shark populations, altering their roles and effects on ecosystems. We review these changes and implications for ecosystem function and management. Macropredatory sharks are often disproportionately affected by humans but can influence prey and coastal ecosystems, including facilitating carbon sequestration. Like terrestrial predators, sharks may be crucial to ecosystem functioning under climate change. However, large ecosystem effects of sharks are not ubiquitous. Increasing human uses of oceans are changing shark roles, necessitating management consideration. Rebuilding key populations and incorporating shark ecological roles, including less obvious ones, into management efforts are critical for retaining sharks' functional value. Coupled social-ecological frameworks can facilitate these efforts.

Promoting fairness in activity recognition algorithms for patient's monitoring and evaluation systems in healthcare.

Researchers face the challenge of defining subject selection criteria when training algorithms for human activity recognition tasks. The ongoing uncertainty revolves around which characteristics should be considered to ensure algorithmic robustness across diverse populations. This study aims to address this challenge by conducting an analysis of heterogeneity in the training data to assess the impact of physical characteristics and soft-biometric attributes on activity recognition performance. The performance of various state-of-the-art deep neural network architectures (tCNN, hybrid-LSTM, Transformer model) processing time-series data using the IntelliRehab (IRDS) dataset was evaluated. By intentionally introducing bias into the training data based on human characteristics, the objective is to identify the characteristics that influence algorithms in motion analysis. Experimental findings reveal that the CNN-LSTM model achieved the highest accuracy, reaching 88%. Moreover, models trained on heterogeneous distributions of disability attributes exhibited notably higher accuracy, reaching 51%, compared to those not considering such factors, which scored an average of 33%. These evaluations underscore the significant influence of subjects' characteristics on activity recognition performance, providing valuable insights into the algorithm's robustness across diverse populations. This study represents a significant step forward in promoting fairness and trustworthiness in artificial intelligence by quantifying representation bias in multi-channel time-series activity recognition data within the healthcare domain.

Elevated levels of antibiotic resistance genes as a factor of human-caused global environmental change.

Antibiotic resistance genes (ARGs) have moved into focus as a critically important response variable in global change biology, given the increasing environmental and human health threat posed by these genes. However, we propose that elevated levels of ARGs should also be considered a factor of global change, not just a response. We provide evidence that elevated levels of ARGs are a global change factor, since this phenomenon is linked to human activity, occurs globally, and affects biota. We explain why ARGs could be considered the global change factor, rather than the organisms containing them; and we highlight the difference between ARGs and the presence of antibiotics, which are not necessarily linked since elevated levels of ARGs are caused by multiple factors. Importantly, shifting the perspective to elevated levels of ARGs as a factor of global change opens new avenues of research, where ARGs can be the experimental treatment. This includes asking questions about how elevated ARG levels interact with other global change factors, or how ARGs influence ecosystem processes, biodiversity or trophic relationships. Global change biology stands to profit from this new framing in terms of capturing more completely the real extent of human impacts on this planet.

A human activity recognition method based on Vision Transformer.

Human activity recognition has a wide range of applications in various fields, such as video surveillance, virtual reality and human-computer intelligent interaction. It has emerged as a significant research area in computer vision. GCN (Graph Convolutional networks) have recently been widely used in these fields and have made great performance. However, there are still some challenges including over-smoothing problem caused by stack graph convolutions and deficient semantics correlation to capture the large movements between time sequences. Vision Transformer (ViT) is utilized in many 2D and 3D image fields and has surprised results. In our work, we propose a novel human activity recognition method based on ViT (HAR-ViT). We integrate enhanced AGCL (eAGCL) in 2s-AGCN to ViT to make it process spatio-temporal data (3D skeleton) and make full use of spatial features. The position encoder module orders the non-sequenced information while the transformer encoder efficiently compresses sequence data features to enhance calculation speed. Human activity recognition is accomplished through multi-layer perceptron (MLP) classifier. Experimental results demonstrate that the proposed method achieves SOTA performance on three extensively used datasets, NTU RGB+D 60, NTU RGB+D 120 and Kinetics-Skeleton 400.

Bio-Skin-Inspired Flexible Pressure Sensor Based on Carbonized Cotton Fabric for Human Activity Monitoring.

With the development of technology, people's demand for pressure sensors with high sensitivity and a wide working range is increasing. An effective way to achieve this goal is simulating human skin. Herein, we propose a facile, low-cost, and reproducible method for preparing a skin-like multi-layer flexible pressure sensor (MFPS) device with high sensitivity (5.51 kPa-1 from 0 to 30 kPa) and wide working pressure range (0-200 kPa) by assembling carbonized fabrics and micro-wrinkle-structured Ag@rGO electrodes layer by layer. In addition, the highly imitated skin structure also provides the device with an extremely short response time (60/90 ms) and stable durability (over 3000 cycles). Importantly, we integrated multiple sensor devices into gloves to monitor finger movements and behaviors. In summary, the skin-like MFPS device has significant potential for real-time monitoring of human activities in the field of flexible wearable electronics and human-machine interaction.

A Robust Deep Feature Extraction Method for Human Activity Recognition Using a Wavelet Based Spectral Visualisation Technique.

Human Activity Recognition (HAR), alongside Ambient Assisted Living (AAL), are integral components of smart homes, sports, surveillance, and investigation activities. To recognize daily activities, researchers are focusing on lightweight, cost-effective, wearable sensor-based technologies as traditional vision-based technologies lack elderly privacy, a fundamental right of every human. However, it is challenging to extract potential features from 1D multi-sensor data. Thus, this research focuses on extracting distinguishable patterns and deep features from spectral images by time-frequency-domain analysis of 1D multi-sensor data. Wearable sensor data, particularly accelerator and gyroscope data, act as input signals of different daily activities, and provide potential information using time-frequency analysis. This potential time series information is mapped into spectral images through a process called use of 'scalograms', derived from the continuous wavelet transform. The deep activity features are extracted from the activity image using deep learning models such as CNN, MobileNetV3, ResNet, and GoogleNet and subsequently classified using a conventional classifier. To validate the proposed model, SisFall and PAMAP2 benchmark datasets are used. Based on the experimental results, this proposed model shows the optimal performance for activity recognition obtaining an accuracy of 98.4% for SisFall and 98.1% for PAMAP2, using Morlet as the mother wavelet with ResNet-101 and a softmax classifier, and outperforms state-of-the-art algorithms.

Human degradation of tropical moist forests is greater than previously estimated.

Tropical forest degradation from selective logging, fire and edge effects is a major driver of carbon and biodiversity loss1-3, with annual rates comparable to those of deforestation4. However, its actual extent and long-term impacts remain uncertain at global tropical scale5. Here we quantify the magnitude and persistence of multiple types of degradation on forest structure by combining satellite remote sensing data on pantropical moist forest cover changes4 with estimates of canopy height and biomass from spaceborne6 light detection and ranging (LiDAR). We estimate that forest height decreases owing to selective logging and fire by 15% and 50%, respectively, with low rates of recovery even after 20 years. Agriculture and road expansion trigger a 20% to 30% reduction in canopy height and biomass at the forest edge, with persistent effects being measurable up to 1.5 km inside the forest. Edge effects encroach on 18% (approximately 206 Mha) of the remaining tropical moist forests, an area more than 200% larger than previously estimated7. Finally, degraded forests with more than 50% canopy loss are significantly more vulnerable to subsequent deforestation. Collectively, our findings call for greater efforts to prevent degradation and protect already degraded forests to meet the conservation pledges made at recent United Nations Climate Change and Biodiversity conferences.

Toward Concurrent Identification of Human Activities with a Single Unifying Neural Network Classification: First Step.

The characterization of human behavior in real-world contexts is critical for developing a comprehensive model of human health. Recent technological advancements have enabled wearables and sensors to passively and unobtrusively record and presumably quantify human behavior. Better understanding human activities in unobtrusive and passive ways is an indispensable tool in understanding the relationship between behavioral determinants of health and diseases. Adult individuals (N = 60) emulated the behaviors of smoking, exercising, eating, and medication (pill) taking in a laboratory setting while equipped with smartwatches that captured accelerometer data. The collected data underwent expert annotation and was used to train a deep neural network integrating convolutional and long short-term memory architectures to effectively segment time series into discrete activities. An average macro-F1 score of at least 85.1 resulted from a rigorous leave-one-subject-out cross-validation procedure conducted across participants. The score indicates the method's high performance and potential for real-world applications, such as identifying health behaviors and informing strategies to influence health. Collectively, we demonstrated the potential of AI and its contributing role to healthcare during the early phases of diagnosis, prognosis, and/or intervention. From predictive analytics to personalized treatment plans, AI has the potential to assist healthcare professionals in making informed decisions, leading to more efficient and tailored patient care.

Floristic diversity and its relationships with human land use varied regionally during the Holocene.

Humans have caused growing levels of ecosystem and diversity changes at a global scale in recent centuries but longer-term diversity trends and how they are affected by human impacts are less well understood. Analysing data from 64,305 pollen samples from 1,763 pollen records revealed substantial community changes (turnover) and reductions in diversity (richness and evenness) in the first ~1,500 to ~4,000 years of the Holocene epoch (starting 11,700 years ago). Turnover and diversity generally increased thereafter, starting ~6,000 to ~1,000 years ago, although the timings, magnitudes and even directions of these changes varied among continents, biomes and sites. Here, modelling these diversity changes, we find that most metrics of biodiversity change are associated with human impacts (anthropogenic land-cover change estimates for the last 8,000 years), often positively but the magnitudes, timings and sometimes directions of associations differed among continents and biomes and sites also varied. Once-forested parts of the world tended to exhibit biodiversity increases while open areas tended to decline. These regionally specific relationships between humans and floristic diversity highlight that human-biodiversity relationships have generated positive diversity responses in some locations and negative responses in others, for over 8,000 years.

Innovative and practical tools for monitoring and assessing biodiversity status and impacts of multiple human pressures in marine systems.

Human activities at sea can produce pressures and cumulative effects on ecosystem components that need to be monitored and assessed in a cost-effective manner. Five Horizon European projects have joined forces to collaboratively increase our knowledge and skills to monitor and assess the ocean in an innovative way, assisting managers and policy-makers in taking decisions to maintain sustainable activities at sea. Here, we present and discuss the status of some methods revised during a summer school, aiming at better management of coasts and seas. We include novel methods to monitor the coastal and ocean waters (e.g. environmental DNA, drones, imaging and artificial intelligence, climate modelling and spatial planning) and innovative tools to assess the status (e.g. cumulative impacts assessment, multiple pressures, Nested Environmental status Assessment Tool (NEAT), ecosystem services assessment or a new unifying approach). As a concluding remark, some of the most important challenges ahead are assessing the pros and cons of novel methods, comparing them with benchmark technologies and integrating these into long-standing time series for data continuity. This requires transition periods and careful planning, which can be covered through an intense collaboration of current and future European projects on marine biodiversity and ecosystem health.

On-Device Semi-Supervised Activity Detection: A New Privacy-Aware Personalized Health Monitoring Approach.

This paper presents an on-device semi-supervised human activity detection system that can learn and predict human activity patterns in real time. The clinical objective is to monitor and detect the unhealthy sedentary lifestyle of a user. The proposed semi-supervised learning (SSL) framework uses sparsely labelled user activity events acquired from Inertial Measurement Unit sensors installed as wearable devices. The proposed cluster-based learning model in this approach is trained with data from the same target user, thus preserving data privacy while providing personalized activity detection services. Two different cluster labelling strategies, namely, population-based and distance-based strategies, are employed to achieve the desired classification performance. The proposed system is shown to be highly accurate and computationally efficient for different algorithmic parameters, which is relevant in the context of limited computing resources on typical wearable devices. Extensive experimentation and simulation study have been conducted on multi-user human activity data from the public domain in order to analyze the trade-off between classification accuracy and computation complexity of the proposed learning paradigm with different algorithmic hyper-parameters. With 4.17 h of training time for 8000 activity episodes, the proposed SSL approach consumes at most 20 KB of CPU memory space, while providing a maximum accuracy of 90% and 100% classification rates.