High Speed and Accuracy of Animation 3D Pose Recognition Based on an Improved Deep Convolution Neural Network

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(13), P. 7566 - 7566

Published: June 27, 2023

Pose recognition in character animations is an important avenue of research computer graphics. However, the current use traditional artificial intelligence algorithms to recognize animation gestures faces hurdles such as low accuracy and speed. Therefore, overcome above problems, this paper proposes a real-time 3D pose system, which includes both facial body poses, based on deep convolutional neural networks further designs single-purpose estimation system. First, we transformed human extracted from input image abstract data structure. Subsequently, generated required at runtime dataset. This challenges conventional concept monocular estimation, extremely difficult achieve. It can also achieve running speed resolution 384 fps. The proposed method was used identify multiple-character using multiple datasets (Microsoft COCO 2014, CMU Panoptic, Human3.6M, JTA). results indicated that improved algorithm performance by approximately 3.5% 8–10 times, respectively, significantly superior other classic algorithms. Furthermore, tested system pose-recognition datasets. attitude reach 24 fps with error 100 mm, considerably less than 2D 60 learning study yielded surprisingly performance, proving deep-learning technology for has great potential.

Language: Английский

---

A cyber-physical robotic mobile fulfillment system in smart manufacturing: The simulation aspect

Robotics and Computer-Integrated Manufacturing, Journal Year: 2023, Volume and Issue: 83, P. 102578 - 102578

Published: April 21, 2023

Language: Английский

---

Sitting Posture Recognition Systems: Comprehensive Literature Review and Analysis

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(18), P. 8557 - 8557

Published: Sept. 23, 2024

Sitting posture recognition systems have gained significant attention due to their potential applications in various domains, including healthcare, ergonomics, and human-computer interaction. This paper presents a comprehensive literature review analysis of existing sitting systems. Through an extensive examination relevant research articles conference papers, we identify analyze the underlying technologies, methodologies, datasets, performance metrics, associated with these The encompasses both traditional methods, such as vision-based approaches sensor-based techniques, well emerging technologies machine learning deep algorithms. Additionally, examine challenges, constraints, future trends field Researchers, practitioners, policymakers who want comprehend most recent developments latest technology will find great value this study.

Language: Английский

---

Evaluation of Convolutional Neural Network-Based Posture Identification Model of Older Adults: From Silhouette of Sagittal Photographs

Geriatrics, Journal Year: 2025, Volume and Issue: 10(2), P. 49 - 49

Published: March 19, 2025

Background/Objectives: Posture is a significant indicator of health status in older adults. This study aimed to develop an automatic posture assessment tool based on sagittal photographs by validating recognition models using convolutional neural networks. Methods: A total 9140 images were collected with data augmentation, and each image was labeled as either Ideal or Non-Ideal physical therapists. The hidden output layers the remained unchanged, while loss function optimizer varied construct four different model configurations: mean squared error Adam (MSE & Adam), stochastic gradient descent SGD), binary cross-entropy (BCE SGD). Results: All demonstrated improved accuracy both training validation phases. However, two BCE exhibited divergence loss, suggesting overfitting. Conversely, MSE showed stability during learning. Therefore, we focused evaluated their reliability sensitivity, specificity, Prevalence-Adjusted Bias-Adjusted Kappa (PABAK) model’s correct label. Sensitivity specificity 85% 84% for 67% 77% SGD, respectively. Moreover, PABAK values agreement label 0.69 0.43 Conclusions: Our findings indicate that model, particular, can serve useful screening inspections.

Language: Английский

---

Multimodal data-based human motion intention prediction using adaptive hybrid deep learning network for movement challenged person

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Dec. 24, 2024

Recently, social demands for a good quality of life have increased among the elderly and disabled people. So, biomedical engineers robotic researchers aimed to fuse these techniques in novel rehabilitation system. Moreover, models utilized signals acquired from human body's particular organ, cells, or tissues. The motion intention prediction mechanism plays an essential role various applications, such as assistive robots, that execute specific tasks elders physically impaired individuals. However, more complications are human–machine-based interaction techniques, creating scope personalized assistance Therefore, this paper, Adaptive Hybrid Network (AHN) is implemented effective prediction. Initially, multimodal data like electroencephalogram (EEG)/Electromyography (EMG) sensor measures collected available resource. gathered EEG/EMG then converted into spectrogram images sent AH-CNN-LSTM, which integration Convolution Neural (AH-CNN) with Long Short-Term Memory (LSTM) network. Similarly, details directly subjected AH-CNN-Res-LSTM, combination CNN Residual LSTM (Res-LSTM) get predictive result. Further, enhance prediction, parameters both AH-CNN-LSTM AH-CNN-Res-LSTM optimized using Improved Yellow Saddle Goatfish Algorithm (IYSGA). efficiency model computed by conducting comparison experiment proposed technique other standard models. performance outcome developed method outperformed traditional methods.

Language: Английский

---

LMSPNet: Improved Lightweight Network for Multi-Person Sitting Posture Recognition

Published: May 26, 2023

Incorrect sitting posture may lead to health problems. Therefore, effective recognition can remind individuals maintain correct and reduce discomfort. Traditional methods for have limitations in terms of high cost slow inference speed. To address these issues, we propose a novel model called LMSPNet multi-person recognition. This first employs the Light Convolution Core (LCC) complexity then introduces Convolutional Block Attention Module (CBAM) adaptively adjust receptive field neural network capture global contextual information, thereby enabling better learn relationships between different channels. We construct human dataset evaluate performance LMSPNet. Experimental results demonstrate that, compared baseline models, our achieves state-of-the-art with an accuracy 99.57%. is expected become powerful tool

Language: Английский

---

Secure Co-Creation of Industrial Knowledge Graph: Graph Complement Method with Federated Learning and ChatGPT

2022 IEEE 18th International Conference on Automation Science and Engineering (CASE), Journal Year: 2023, Volume and Issue: unknown, P. 1 - 6

Published: Aug. 26, 2023

Industrial areas have increasingly developed their own Knowledge Graph (KG) for organizing and leveraging vast amounts of data. One major challenge in constructing KG is the heavy reliance on available resources, restricting scalability accuracy resulting graphs. To address this issue, an end-to-end method proposed to create a multi-benefit ecosystem by integrating Federated Learning with ChatGPT (a popular language model). Different stakeholders may leverage search novel knowledge that complements existing KGs, however, approach could potentially introduce ambiguous wrong triples into KG. overcome this, applied align disambiguate using other industrial KGs as super-vision. The applies multi-field hyperbolic embedding vectorize entities edges, which are then associatively aggregated achieve edge replenishment entity fusion each encrypted. Finally, incentive win-win mechanism motivate diverse contribute co-creation actively. A case study conducted different evaluate method. Results demonstrate provides practical solution no compromise data security.

Language: Английский

---

Analysis and Improvement of the Application of Playground Sports Posture Detection Technology in Physical Education Teaching and Training

EAI Endorsed Transactions on Pervasive Health and Technology, Journal Year: 2024, Volume and Issue: 10

Published: March 18, 2024

INTORDUCTION: The goal of human posture detection technology applied in the field sports is to realise indexing norms, provide scientific guidance for training and teaching, which great significance improve quality sports.OBJECITVES: Aiming at problems incomplete features, low accuracy real-time performance recognition methods.METHODS: In this paper, a method pose based on snow melting heuristic optimisation algorithm deep limit learning machine network proposed. Firstly, by analyzing process motion detection, extracting feature coordinates Blaze-Pose Blaze-Hands key nodes, constructing system; then, optimizing parameters extreme through snow-melt optimization algorithm, model; finally, simulation experiments analysis, proposed method's can reach 95% time less than 0.01 s.RESULTS: results show that improves precision, robustness performance.CONCLUSION: problem poor generalisation, insufficient application solved.

Language: Английский

---

A method for complex posture recognition during long-term sitting using neural networks and pressure mapping systems

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 95, P. 106306 - 106306

Published: April 17, 2024

This study aimed to identify correct sitting postures during prolonged periods using a pressure mapping system on the seat along with Convolutional Neural Networks (CNN). The was conducted in three stages. In first stage, twenty-two volunteers participated obtain dataset of maps (flat, short lordosis, and long lordosis) two systems validating methods. involved angle measurement through an Inertial Measurement Unit (IMU) image recognition. second CNN model trained data from stage that represented each posture, then Transfer Learning implemented different system. third used for long-term monitoring based model, feedback number provided healthy individuals at least 2 h. posture recognition participants. Additionally, participants evaluated their experience Likert-5 questionnaire. results showed accurately identified accuracy 0.854, precision 0.856, recall 0.854. found collection methodology non-invasive unobtrusive. Moreover, system's understandable helpful improving posture. approach presented this has potential facilitate further research since it can be easily adapted various without altering methodology.

Language: Английский

---

Modeling the Impact of Sitting Posture and Duration on the Health and Safety of Remote Learning

Published: Feb. 12, 2024

This study investigates the effect of sitting posture on back health students involved in remote work. Jack® human simulation software was employed to measure impact different seated postures lower region, specifically lumbar spine. The analysis focused common positions adopted during extended work periods: neutral (Posture 1), flexion 2) and extension 3). Three tests were conducted simulated environment predict forces exerted musculoskeletal system, which are Lower (LBA), Comfort Analysis, Ovako Working Posture Analysis System (OWAS). results revealed that 2 exhibited highest compression force is least comfortable terms joint angles compared other while 3 body parts. Furthermore, loading associated with working does not reach excessive levels, however it required provide ergonomic improvements. paper also proposes strategies enhance university students.

Language: Английский

---