Activity Recognition System Research Articles

SkelVIT: consensus of vision transformers for a lightweight skeleton-based action recognition system

SkelVIT: consensus of vision transformers for a lightweight skeleton-based action recognition system

Humanactivity Recognition using Deep Learning

Human activity recognition (HAR) plays a crucial role in various fields, including healthcare, surveillance, and human-computer interaction. This study explores the application of deep learning techniques for accurate and efficient human activity recognition. Leveraging the capabilities of deep neural networks, specifically convolutional neural networks (CNNs) and recurrent neural networks (RNNs), the proposed approach aims to capture both spatial and temporal features from sensor data. The dataset utilized in this research comprises multi-model sensor inputs, such as accelerometer and gyroscope readings, collected from wearable devices. The deep learning model is designed to automatically learn hierarchical representations of the raw sensor data, enabling robust feature extraction and discrimination between different human activities. Transfer learning is employed to enhance model generalization across diverse activity categories and varying sensor setups. Experimental evaluations are conducted on real-world datasets, demonstrating the effectiveness of the proposed deep learning framework in accurately classifying activities, including walking, running, sitting, and standing. Comparative analyses against traditional machine learning methods underscore the superior performance and adaptability of deep learning in handling complex and dynamic activity patterns. The results showcase the potential of deploying deep learning models in real-time human activity recognition systems, highlighting their scalability and efficiency. The study contributes to the advancement of HAR methodologies, paving the way for the development of more reliable and robust systems in applications such as healthcare monitoring, assisted living, and smart environments.

Hybrid Approach Using Machine Learning and IOT for Soldier Rescue : A Review

Despite the well-established influence of various factors on a soldier's burden – including environment, physical exertion, equipment design, and mental stress – our current understanding is largely based on studies conducted in controlled lab settings, focusing solely on the weight of carried equipment. This limited scope hinders a comprehensive picture of how these combined burdens impact a soldier's ability to survive on the battlefield, encompassing factors like performance, health, and vulnerability to enemy attacks. To bridge this gap and gain a more holistic understanding, field-based methods for capturing soldier movement are crucial. In this vein, we've developed a novel human activity recognition system. Trained using data collected from a single sensor placed on a soldier's upper back, the system can identify eleven distinct tactical movement patterns commonly employed by soldiers in the field. This advancement paves the way for a more nuanced understanding of how various burden factors interact and influence a soldier's effectiveness and safety in real-world scenarios. Using K- Nearest Neighbour, SVM Classifier, Logistic Regression, Naïve Bayer algorithms real-world constraints are forced, and class labels are expanded. This project is based on health monitoring and tracking system for soldiers. The proposed system can be mounted on the soldier’s jacket to track their health status and current locating using GPS. This information will be transmitted to the control room through IOT and ML. The proposed system comprises of tiny wearable physiological equipment’s, sensors, transmission modules. Hence, with the use of the proposed equipment, it is possible to implement a low-cost mechanism to protect the valuable human life on the battlefield. It also includes about securing of data of soldiers in the cloud. This new method offers a powerful tool for military leaders and scientists. By collecting real-world data on soldier burden, it allows them to quantify the complex factors affecting soldier performance (the tradespace). This data acts as valuable pre-processing for other technologies, ultimately enabling data-driven decisions to optimize soldier well-being, minimize risk, and maximize mission success.

Human Action Recognition using Long Short-Term Memory and Convolutional Neural Network Model

Human Action Recognition (HAR) is the difficulty of quickly identifying strenuous exercise performed by people. It is feasible to sample some measures of a body's tangential acceleration and speed using inertial sensors and exercise them only to learn model skills of incorrectly categorizing behavior into the relevant categories. In detecting human activities, the use of detectors in personal and portable devices has increased to better understand and anticipate human behavior. Many specialists are working toward developing a classification that can distinguish between a user's behavior and uncooked data while utilizing as few reserves as possible. A Long-term Recurrent Convolutional Network (LRCN) is proposed as a comprehensive human action recognition system based on deep neural networks in this paper.

WiCAR: A class-incremental system for WiFi activity recognition

The proposal of Integrated Sensing and Communications has once again drawn researchers’ attention to WiFi sensing, propelling applications based on WiFi sensing into an advanced stage. However, the current field of activity recognition only identifies fixed categories of activities, neglecting the growing demand for perceiving activity types in real applications over time. In response to the issue, we present WiCAR, a WiFi activity recognition system designed for class incremental scenarios. WiCAR takes antenna array-fused image data as input, employing the Wi-RA model with parallel stacked activation functions as its backbone network. To alleviate the typical catastrophic forgetting issue in class-incremental learning, WiCAR employs a strategy of replaying known data. Additionally, we adopts knowledge distillation to improve accuracy among old samples during the incremental process. To tackle the imbalance in the number of samples between old and new classes, the model is updated through weight alignment. This serious of strategies endows the system with the capability to progressively learn and handle new classes. We conducted extensive experiments to evaluate the system performance. The experimental results demonstrate that our system exhibits excellent performance regardless of the number of tasks, whether tasks are uniform or non-uniform, and the order of task arrivals. The highest average accuracy reaches 96.429%, and even in the presence of six incremental stages, the average accuracy remains at 92.867%.

Resource-efficient, sensor-based human activity recognition with lightweight deep models boosted with attention

With their automatic feature extraction capabilities, deep learning models have become more widespread in sensor-based human activity recognition, particularly on larger datasets. However, their direct use on mobile and wearable devices is challenging due to the extensive resource requirements. Concurrently, attention-based models are emerging to improve recognition performance by dynamically emphasizing relevant parts of features and disregarding the irrelevant ones, particularly in the computer vision domain. This study introduces a novel application of attention mechanisms to smaller deep architectures, investigating whether smaller models can achieve comparable recognition performance to larger models in sensor-based human activity recognition systems while keeping resource usage at lower levels. For this purpose, we integrate the convolutional block attention module into a hybrid model, deep convolutional and long short-term memory network. Experiments are conducted using five public datasets in three model sizes: lightweight, moderate and original. The results show that applying attention to the lightweight model enables achieving similar recognition performances to the moderate-size model, and the lightweight model requires approximately 2–13 times fewer parameters and 3.5 times fewer flops. We also conduct experiments with sensor data at lower sampling rates and from fewer sensors attached to different body parts. The results show that attention improves recognition performance under lower sampling rates, as well as under higher sampling rates when model sizes are smaller, and mitigates the impact of missing data from one or more body parts, making the model more suitable for real-world sensor-based applications.

The Deep Learning-based Human Action Recognition System for Competitive Sports

The Deep Learning-based Human Action Recognition System for Competitive Sports

The Deep Learning-based Human Action Recognition System for Competitive Sports

The Deep Learning-based Human Action Recognition System for Competitive Sports

Privacy-Preserving Human Activity Recognition System for Assisted Living Environments

Privacy-Preserving Human Activity Recognition System for Assisted Living Environments

Fall Detection System based on iOS Smartphone Sensors

Activities of daily living (ADLs) are a crucial aspect of human life, especially in remote health monitoring and fall detection. As smartphones have become an integral part of our daily routines, with their ability to perform complex calculations, connect to the internet, and incorporate various sensors, researchers have been inspired to explore human activity recognition systems. This paper focuses on accelerometer and gyroscope data from iOS-based smartphones. We developed a data collection app to record fall types (e.g., Falling Right, Falling Left) and fall-like activities (e.g., Sitting Fast, Jumping). Volunteers carried smartphones naturally in their pockets during experiments, presenting a challenge of noise but enhancing user comfort. We applied different Machine Learning algorithms (Decision Trees, Random Forest, Logistic Regression, k-Nearest Neighbor, XGBoost, LightGBM, and Neural Networks) to analyze the collected dataset. In contrast to typical studies, our approach replicated real-world smartphone usage. The paper presents and analyzes promising results from the study. Furthermore, we implemented the trained model as a real-time mobile application for potential users. This research illustrates the potential of smartphones in fall detection and opens the way for user-friendly solutions in remote health monitoring.

Human activity recognition using a single-photon direct time-of-flight sensor

Single-Photon Avalanche Diode (SPAD) direct Time-of-Flight (dToF) sensors provide depth imaging over long distances, enabling the detection of objects even in the absence of contrast in colour or texture. However, distant objects are represented by just a few pixels and are subject to noise from solar interference, limiting the applicability of existing computer vision techniques for high-level scene interpretation. We present a new SPAD-based vision system for human activity recognition, based on convolutional and recurrent neural networks, which is trained entirely on synthetic data. In tests using real data from a 64×32 pixel SPAD, captured over a distance of 40 m, the scheme successfully overcomes the limited transverse resolution (in which human limbs are approximately one pixel across), achieving an average accuracy of 89% in distinguishing between seven different activities. The approach analyses continuous streams of video-rate depth data at a maximal rate of 66 FPS when executed on a GPU, making it well-suited for real-time applications such as surveillance or situational awareness in autonomous systems.

Novel hybrid optimization based adaptive deep convolution neural network approach for human activity recognition system

Novel hybrid optimization based adaptive deep convolution neural network approach for human activity recognition system

Human Activity Recognition Based on Deep Learning and Micro-Doppler Radar Data.

Activity recognition is one of the significant technologies accompanying the development of the Internet of Things (IoT). It can help in recording daily life activities or reporting emergencies, thus improving the user's quality of life and safety, and even easing the workload of caregivers. This study proposes a human activity recognition (HAR) system based on activity data obtained via the micro-Doppler effect, combining a two-stream one-dimensional convolutional neural network (1D-CNN) with a bidirectional gated recurrent unit (BiGRU). Initially, radar sensor data are used to generate information related to time and frequency responses using short-time Fourier transform (STFT). Subsequently, the magnitudes and phase values are calculated and fed into the 1D-CNN and Bi-GRU models to extract spatial and temporal features for subsequent model training and activity recognition. Additionally, we propose a simple cross-channel operation (CCO) to facilitate the exchange of magnitude and phase features between parallel convolutional layers. An open dataset collected through radar, named Rad-HAR, is employed for model training and performance evaluation. Experimental results demonstrate that the proposed 1D-CNN+CCO-BiGRU model demonstrated superior performance, achieving an impressive accuracy rate of 98.2%. This outperformance of existing systems with the radar sensor underscores the proposed model's potential applicability in real-world scenarios, marking a significant advancement in the field of HAR within the IoT framework.

A Simulation-Based Framework for the Design of Human Activity Recognition Systems Using Radar Sensors

A Simulation-Based Framework for the Design of Human Activity Recognition Systems Using Radar Sensors

Wi-CHAR: A WiFi Sensing Approach with Focus on Both Scenes and Restricted Data.

Significant strides have been made in the field of WiFi-based human activity recognition, yet recent wireless sensing methodologies still grapple with the reliance on copious amounts of data. When assessed in unfamiliar domains, the majority of models experience a decline in accuracy. To address this challenge, this study introduces Wi-CHAR, a novel few-shot learning-based cross-domain activity recognition system. Wi-CHAR is meticulously designed to tackle both the intricacies of specific sensing environments and pertinent data-related issues. Initially, Wi-CHAR employs a dynamic selection methodology for sensing devices, tailored to mitigate the diminished sensing capabilities observed in specific regions within a multi-WiFi sensor device ecosystem, thereby augmenting the fidelity of sensing data. Subsequent refinement involves the utilization of the MF-DBSCAN clustering algorithm iteratively, enabling the rectification of anomalies and enhancing the quality of subsequent behavior recognition processes. Furthermore, the Re-PN module is consistently engaged, dynamically adjusting feature prototype weights to facilitate cross-domain activity sensing in scenarios with limited sample data, effectively distinguishing between accurate and noisy data samples, thus streamlining the identification of new users and environments. The experimental results show that the average accuracy is more than 93% (five-shot) in various scenarios. Even in cases where the target domain has fewer data samples, better cross-domain results can be achieved. Notably, evaluation on publicly available datasets, WiAR and Widar 3.0, corroborates Wi-CHAR's robust performance, boasting accuracy rates of 89.7% and 92.5%, respectively. In summary, Wi-CHAR delivers recognition outcomes on par with state-of-the-art methodologies, meticulously tailored to accommodate specific sensing environments and data constraints.

Hybrid HAR-CNN Model: A Hybrid Convolutional Neural Network Model for Predicting and Recognizing the Human Activity Recognition

Human activity recognition (HAR) is an active research area in computer vision from past several years and research is still continuing in this field due to the unavailability of perfect recognition system. The human activity recognition system it covers e-health, patient monitoring, assistive daily living activities, video surveillance, security and behaviour analysis, and sports analysis. Many researchers have suggested techniques that use visual perception to detect human activities. Researchers will need to address problems including light variations in human activity detection, interclass similarity between scenes, the surroundings and recording setting, and temporal variation in order to construct an efficient vision-based human activity recognition system. However, a significant drawback of many deep learning models is their inability to achieve satisfactory results in real-world scenarios due to the conflicts mentioned above. To address this challenge, we developed a hybrid HAR-CNN classifier aimed at enhancing the learning outcomes of Deep CNNs by combining two models: Improved CNN and VGG-19. Using the KTH dataset, we collected 6,000 images for training, validation, and testing of our proposed technique. Our research findings indicate that the Hybrid HAR-CNN model, which combines Improved CNN with VGG-19 Net, outperforms individual deep learning models such as Improved CNN and VGG-19 Net.

Exploiting channel state information of WiFi signal for human activity detection: an experimental study

Ubiquitous computing aims to seamlessly integrate computing into our daily lives, and requires reliable information on human activities and state for various applications. In this paper, we propose a device-free human activity recognition system that leverages the rich information behind WiFi signals to detect human activities in indoor environments, including walking, sitting, and standing. The key idea of our system is to use the dynamic features of activities, which we carefully examine and analyze through the characteristics of channel state information. We evaluate the impact of location changes on WiFi signal distribution for different activities and design an activity detection system that employs signal processing techniques to extract discriminative features from wireless signals in the frequency and temporal domains. We implement our system on a single off-the-shelf WiFi device connecting to a commercial wireless access point and evaluate it in laboratory and conference room environments. Our experiments demonstrate the feasibility of using WiFi signals for device-free human activity recognition, which could provide a practical and non-intrusive solution for indoor monitoring and ubiquitous computing applications.

Enhanced Sign Language Communication through Deep Learning-based Gesture and Action Recognition and Correction

Sign language is a crucial form of communication for the hearing impaired, presenting unique challenges for technological interpretation. This project focuses on developing a Sign Language Detection Using Action Recognition (SLDAR) system to address these challenges. Leveraging Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and multimodal fusion techniques, our system aims to accurately detect and recognize sign language gestures in real-time. The project involves collecting and annotating a diverse dataset comprising video, audio, and text data. We then design and train a deep learning model that combines features from these modalities to improve accuracy and robustness.The project involves collecting and annotating a diverse dataset, designing a deep learning model for multimodal fusion, and implementing a user-friendly interface for real-time interaction. The system's performance is evaluated using metrics such as accuracy, precision, and recall. Our goal is to enhance accessibility and communication for the hearing impaired through the development of this advanced sign language recognition system. Keywords:Acoustic Features, Linguistic Context, Mean Opinion Score (MOS), Natural Language Processing, NLP Algorithms, Signal-to-Noise Ratio (SNR), Speech Enhancement, Speech Quality, Telecommunication Systems, Voice Assistants.

Stretchable Electronic Skin using Laser‐Induced Graphene and Liquid Metal with an Action Recognition System Powered by Machine Learning

AbstractMonitoring tactile pressure and recognizing action are important functionalities for artificial electronic skin (e‐skin). Furthermore, in order to create conformable coverings for 3D objects, an e‐skin needs to be stretchable, without sacrificing sensitivity to tactile pressure. However, stretching of sensors normally affects their output stability. In this study, a stretchable e‐skin is developed using laser‐induced graphene and a liquid metal alloy, GaInSn, in an elastic ecoflex polymer to create a stretchable, resistive‐type tactile pressure sensor. Furthermore, a pressure sensor array is fabricated as an e‐skin, and output is signal‐processed using machine learning. With this system, the e‐skin also monitors its stretching state, with the result that tactile pressure can be calculated regardless of the degree of stretching. With machine learning‐assisted e‐skin, actions such as patting, sliding, and grabbing are successfully recognized in the manner of human skin.

Multi-sensor fusion based optimized deep convolutional neural network for boxing punch activity recognition

Recent advancements in deep learning have significantly enhanced the recognition of player activities in sports by enabling automatic feature extraction. In our proposed work, we focus on recognizing six distinct punches in the context of boxing. We incorporate the sliding window technique during the pre-processing stage to transform the time-series data obtained from Inertial Measurement Unit (IMU) sensors in a boxing punch activity recognition system. Our approach influences a sensor fusion-based Deep Convolutional Neural Network (DCNN) classification model to identify various boxing punches accurately, achieving an impressive F1 score. The system demonstrates proficiency in distinguishing similar activities, such as jab and hook punches where the existing systems made misclassifications due to subtle variations in arm flexion that differentiate the two. Through experimentation, we identify an optimal window size for boxing punch activity recognition, which falls within the range of 15–20 frames (equivalent to 0.15–0.25 s). This window size selection results in a notable reduction in inference time. To evaluate our proposed model, we conduct comparisons with a standard DCNN and an optimized DCNN model. Our proposed optimized DCNN model demonstrates enhanced recognition accuracy, achieving an impressive 99%, coupled with an improved F1 score of 87%. Furthermore, the model displays a remarkable reduction in inference time, clocking in at less than 1 ms. Overall, our research contributes to the field of sports-related player activity recognition by employing the power of deep learning. By expertly combining these techniques, we achieve remarkable accuracy, precision, and efficiency in recognizing various boxing punches.