Publications
Parkinson’s Disease Auxiliary Diagnosis System Based on Human Activity Recognition Using Stacked LSTM and GRU Networks
Sep 30, 2025Journal Journal of Wireless Mobile Networks, Ubiquitous Computing, and Dependable Applications
DOI 10.58346/JOWUA.2025.I3.028
Issue 3
Volume 16
Parkinson’s disease (PD) is a progressive neurodegenerative disorder that remains difficult to diagnose at an early stage, as most clinical symptoms appear only after substantial neuronal loss. This study proposes an integrated deep learning based auxiliary diagnosis system for PD using human gait data collected from wearable triaxial accelerometer sensors. The system employs stacked Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks to capture complex temporal dependencies and motor anomalies associated with Parkinsonian gait. Experiments were conducted on the DeFOG dataset, which consists of multi-subject gait recordings annotated with freezing of gait (FoG) events. After rigorous preprocessing, including data cleaning, normalization, and sliding-window segmentation, the stacked GRU model achieved superior results, with 95.9% accuracy, 90.7% precision, 83.6% recall, 87.0% F1-score, and an AUC-ROC of 0.92. These results significantly outperformed traditional machine learning baselines such as Support Vector Machines (SVM) at 86.2%, Random Forests (RF) at 84.7%, and k-Nearest Neighbors (KNN) at 82.9%, as well as deep learning baselines, including CNN-only 90.1% and LSTM-only 92.5% models. Ablation studies confirmed the critical role of preprocessing and multi-layer temporal modeling in improving classification performance. With a low inference latency of 8 ms/sample, the system is well-suited for real-time deployment on wearable devices. While challenges remain in detecting brief FoG episodes and mitigating sensor variability, these results demonstrate that a stacked LSTM–GRU motion analysis system offers reliable, non-invasive support for early PD detection and continuous monitoring in real-world clinical and home settings.
Human Activity-Based Machine Learning and Deep Learning Techniques
Mar 20, 2025DOI 10.22399/ijcesen.1368
Human activity recognition (HAR) has been hot research issues in recent years. The studies have differences in data types, data processing, feature description, etc. HAR constitutes a fundamental component of intelligent health monitoring systems, wherein the underlying intelligence of the services is derived from and enhanced by sensor data. Researchers have proposed multiple HAR systems designed to convert smartphone readings into other forms of physical activity. This review synthesizes the current methodologies for smartphone-based Human Activity Recognition (HAR) with focusing on healthcare application. For this purpose, we systematically searched for peer-reviewed articles regarding the utilization of cell phones for Human Activity Recognition (HAR). We collect information regarding smartphone body placement, sensors, types of physical activities examined, as well as the data transformation methodologies and classification frameworks employed for activity recognition. Thus, we selected these articles and delineated the diverse methodologies employed for data gathering, preprocessing, extraction of features, and activity classification, highlighting the predominant practices and their alternatives. We determine that cell phones are very appropriate for HAR research within the health sciences. Future studies should prioritize enhancing the quality of data gathered, addressing data gaps, incorporating a more diverse array of participants and activities, relaxing phone placement requirements, providing comprehensive documentation for study participants, and sharing the source code of the employed methods and algorithms to achieve population-level impact. © IJCESEN.
Association Rules and Deep Learning Paradigms for Big Data Processing
Nov 1, 2024Journal Journal of Prospective Researches
Volume 24
This in-depth study looks at data on fuel use in two main ways to find trends and make better predictions. One way is to learn how to use machine learning and association rule mining to try to guess what will happen. It uses association rules to show how things in a set are linked and how they do a lot of different things together. We can learn more about the parts that work together to change how much fuel is used. The RNN, TCN, and LSTM machine learning models can all guess how much fuel will be used, but they can do so in different ways. The TCN plan works out the best. The results show how important it is to choose a model design that makes the dataset's features better by putting together numbers and people's ideas about what might be important. We might be able to fully understand how fuel use changes over time if we put together what machine learning and association rule mining tell us. The numbers make it clear that the collection should be used for more research. There are different sets of methods that should be used for machine learning and more general statistical methods. People who give money, make rules, and try to guess how people will use fuel in the future should think about these ideas. It was found that people will be able to make better predictions in the future if they learn more about complicated machine learning design and link rules. The study is a good way to find out how much fuel people use when they switch sources of energy. When people use different types of fuel, we can also see how much they use. We can make the most of what we have this way.
Network Fortification: Leveraging Support Vector Machine for Enhanced Security in Wireless Body Area Networks
Jun 3, 2024Journal International Journal of Safety and Security Engineering
Publisher International Information and Engineering Technology Association
DOI 10.18280/ijsse.140323
This study focuses on enhancing security in wireless body area networks (WBANs) through the application of Support Vector Machine (SVM)-based anomaly detection. The main problem addressed is the insufficient attention to security measures in WBANs, particularly in terms of secure connections and mitigation strategies. The proposed solution involves utilizing SVM to categorize security measures for WBAN telehealth solutions based on relevant attributes, ensuring ongoing utilization. The primary results showcase the successful prediction of vital signs with a remarkable accuracy of 98.63% using SVM, highlighting its effectiveness in enhancing security in WBANs. This paper explores the application of Support Vector Machines (SVMs) to enhance WBAN security updates and intelligence. Specific access management approaches may prove more effective during crisis situations. This study categorizes security measures for WBAN telehealth solutions exclusively using SVM based on relevant security attributes, ensuring their ongoing utilization. Employing SVM, the study predicts a heart rate of 89.087 beats per minute, an RR interval of 673.5 ms, and a QT interval of 271.3 ms, achieving a remarkable accuracy of 98.63 percent with a training dataset comprising 80 percent of the data and a testing dataset encompassing the remaining 20 percent. ©2024 The authors.
Design of A Dual-band Square Slot Antenna for GPS Applications
Aug 1, 2022Journal TELEMATIQUE
Issue 1, 2022
Volume 21
A New Design single feed dual band micro-strip patch antenna that implemented in GPS applications is presented. The suggested antenna based on two bands of GPS systems. they are L1=1.575 GHz and L2=1.227 GHz. The suggested antenna principally comprises two resonance frequencies with contents square patch, two semicircle , two rectangular slots, S-shaped slot, two truncated slot and four circle slots in the ground plane, which fed by a 50 coaxial probe, with over all area of 94*94 mm2. The software CST Microwave studio is used to build a simulation for the suggested antenna. The results of the simulation showed an impedance for bandwidth of 20 MHz between (1.56 GHz-1.58 GHz) by using L1 band, while L2 band, it was 20 MHz between (1.21-1.23GHz). The results also showed that the circular polarization radiation can be obtained through a probe feeding at one of the diagonal lines of the patch. This antenna can reduce the loss in S11, impedance behavior, Gain and farfield radiation pattern.
Medical Data Technology For Automatic Diagnosis System
Jan 1, 2022Journal Journal of Positive Sciences (JPS)
Issue 12
Volume 2022
New industries utilising image processing technology in the medical business, notably for sickness detection, are emerging as a result of the popularity of image processing and its benefits. Grayscale images, such as X-rays and CT scans, provide a classification problem in medical applications since only grey channel information is available and no chronic information is available. A convolutional neural network (CNN) is utilised to detect breast cancer in this study, with huge coloured MRI images used to train the CNN model. Other models used were Random Forest, K-nearest Neighbour, and Nave Biase. CNN's cancer prediction accuracy of 95.12 percent may be maintained in the future.
Analysis of hydraulic characteristics for hollow semi-circular weirs using artificial neural networks
May 21, 2014Journal Flow Measurement and Instrumentation
Publisher Elsevier
DOI 10.1016/j.flowmeasinst.2014.05.003
Volume 38
Weirs are small overflow dams used to alter and raise water flow upstream and regulate or spill water downstream watercourses and rivers. This paper presents the application of artificial neural network (ANN) to determine the discharge coefficient (Cd) for a hollow semi-circular crested weirs. Eighty five experiments were performed in a horizontal rectangular channel of 10 m length, 0.3 m width and 0.45 m depth for a wide range of discharge. The results of examination for discharge coefficient were yielded by using multiple regression equation based on dimensional analysis. Then, the results obtained were also compared using ANN techniques. A multilayer perceptron MLP algorithm FFBP network was developed. The optimal configuration of ANN was [2,10,1] which gave mean square error (MSE) and correlation coefficient (R) of 0.0011 and 0.91, respectively. Performances of ANN model reveal that the Cd could be better estimated by the ANN technique in comparison with Cd obtained using statistical approach.