https://doi.org/10.5573/IEIESPC.2024.13.4.313

(Md Allmamun) ; (Fahima Akter) ; (Muhammad Borhan Uddin Talukdar) ; (Sovon Chakraborty) ; (Jia Uddin)

Drones are flying objects that may be controlled remotely or programmed to do a wide range of tasks, including aerial photography, videography, surveys, crop and animal monitoring, search and rescue missions, package delivery, and military operations. Unrestrained use, however, can pose a significant threat to safety, privacy, and security through eavesdropping, flying close to prohibited locations, interfering with public events, and delivering illicit items. Hence, real-time drone detection and tracking are indispensable and appropriate measures. This study developed real-time drone detection and tracking using the most efficient deep-learning approaches. The models were fine-tuned first to suit the required purpose and yield the desired outcome. The performance of the developed system was better than that of earlier endeavors in terms of accuracy and loss. Of the seven fined-tuned models, the Xception model constantly rendered the maximum accuracy with negligible loss. The model outperformed other state-of-the-art architectures, exhibiting an accuracy and loss of 99.18% and 3.83, respectively.

https://doi.org/10.5573/IEIESPC.2024.13.4.322

(Shengnan Liu) ; (Xu Wang)

Precise extraction of the main melody from polyphonic music is a critical challenge in vocal information processing. This paper starts with a brief introduction to extracting vocal music information features. Two distinct feature types were selected: the Mel-frequency cepstral coefficient (MFCC) and chroma. An innovative main melody extraction algorithm was then developed using a convolutional neural network (CNN) and conditional random field (CRF). The performance of the algorithm was validated on datasets. The main melody extraction effects were improved significantly using MFCC and chroma as inputs to the CNN-CRF algorithm for feature extraction. The algorithm achieved an overall accuracy (OA) of 86.72% and a voicing false alarm (VFA) of 6.84% on the ADC2004 dataset. On the MIREX05 dataset, the algorithm attained an OA and VFA of 85.21% and 11.16%, respectively. The algorithm exhibited pronounced enhancement when being tested on the MIREX05 dataset, and chroma played a notable role in enhancing the raw chroma accuracy. This algorithm also performed better than the SegNet and FTANet algorithms.

https://doi.org/10.5573/IEIESPC.2024.13.4.328

(Bilal Ahmed Lodhi) ; (Rehmat Ullah) ; (Sajida Imran) ; (Muhammad Imran) ; (Byung-Seo Kim)

This paper presents SenseNet, a convolution neural network (CNN) model for image segmentation. SenseNet architecture includes encoders with their corresponding decoders and bottleneck skip connections. The last layer of the architecture is a classification layer that classifies each pixel of an image for image segmentation. SenseNet addresses the limitations of conventional semantic segmentation models. Moreover, the skip connection does not include sufficient information for the recovery in the decoder path. This paper proposes a novel network structure combining a modified dense block and dense skip connection for efficient information recovery at the decoder path. Furthermore, this paper also proposes a dense, long skip connection that transfers the feature maps of each layer of the encoder to a layer of the decoder. This dense skip connection helps the network recover the information efficiently in the decoder path. SenseNet achieves state-of-the-art accuracy with fewer parameters and high-level features in the decoder path. This study evaluated SenseNet on the urban scene benchmark dataset CamVid and measured the performance in terms of intersection over union (IoU) and global accuracy. SenseNet outperformed the baseline model by an 8.7% increase in IoU. SenseNet can be downloaded from https://github.com/sensenetskip/sensenet.

https://doi.org/10.5573/IEIESPC.2024.13.4.337

(Dong-yeon Shin) ; (Seong-won Lee)

In this paper, we track swimmers in an image dataset by using the multi-object tracking (MOT) network called FairMOT. There are problems when tracking swimming datasets through this network, and to address those problems we added a scene detection network to classify input images and to adjust optimized weights and hyper-parameters according to the classes. In addition to the existing IoU formula, we improved performance by assigning optimized weights to each class based on the aspect ratio, considering the center point distance and the aspect ratio between bounding boxes. Additionally, we found optimized values for the hyper-parameters, position, and velocity of the Kalman filter for each class. Utilizing the classifications by the scene detection network, we adjusted the optimized hyper-parameters based on the scene, resulting in a maximum improvement of 5.2% in multi-object tracking accuracy scores.

https://doi.org/10.5573/IEIESPC.2024.13.4.344

(Prabhat Singh) ; (Shilpi Sharma) ; (Ahmed E. Kamal) ; (Sunil Kumar)

Roads are an important part of transporting goods and products from one place to another. In developing countries, the main challenge is to maintain road conditions regularly. Roads can deteriorate from time to time. Monitoring the conditions of the roads, which may degrade with time, is very difficult, resulting in a delay in transportation and damage to the vehicles moving on the roads. Poor road conditions cause road accidents. A model is being proposed to monitor the conditions of the road surface by smartphone sensors. Accelerometer, gyroscope, and GPS sensors are deployed in the mobile phones, which will help to collect data on the road conditions. After collecting the data about the road conditions, various machine learning approaches, such as supervised, multi-layered, and multiclass, are applied to data filtration. Road conditions are divided into three categories to achieve this methodology: potholes, deep traverse cracks, and smooth roads. This categorization helped in analyzing the road surface condition through smartphone sensors over all three axes instead of taking it over a single axis. Neural networks helped analyze data or road conditions more accurately than Decision Tree and SVM.

https://doi.org/10.5573/IEIESPC.2024.13.4.354

(Subhash Mondal) ; (Mithun Karmakar) ; (Amitava Nag)

It has been generally observed that a set of clinically established features can be used to predict obesity. Due to lifestyle habits, most of the population deviates from the suggested treatment to control the state of obesity. This study is an experimental analysis of the effect of related features on the classification of obesity. Two research questions have been designed: “With what degree of accuracy can obesity be categorized using a feature vector (FS) with 16 features?” (RQ1), and, “Can a feature subset (FSS) classify the disease with an accuracy of over 90% compared to the accuracy obtained in RQ1?” (RQ2). It was observed that an FS comprising 16 features reflected an accuracy of 96.68% in the classification of obesity in RQ1, and an FSS comprising four features (selected using the SelectKBest algorithm) exhibited an accuracy of 88.38% on the same dataset. Since 88.38% is 91.42% of 96.68%, the FSS attains accuracy over 90% concerning FS in classifying obesity. Three machine learning (ML) models were selected based on the best accuracy values in the literature. Moreover, both RQ1 and RQ2 have far better accuracy than other methods.

https://doi.org/10.5573/IEIESPC.2024.13.4.361

(Haiying Qi)

The online evaluation data of an image of tourist attractions is conducive to tourists’ objective and fair perception of the destination. This study built a tourist perception model based on online evaluation data of an image of Xi’an tourist attractions. The model first uses the TF-IDF algorithm to analyze the cognitive image of tourists. It then uses the NB method to analyze the emotional image of tourists, and finally, it uses an LDA theme model to analyze the overall image of the scenic spot to explore the tourist perception. The range of TF-IDF values is 0.0245-0.2316, and the maximum value and minimum value correspond to the service attitude and category, respectively. The NB model has a long running time under different data scales, and the corresponding maximum values are 8.1 s and 7.9 s. With the same data size, NN has the shortest running time, followed by SVM and KNN. When the number of topics is 4, the confusion degree of positive emotional text and negative emotional text are the lowest, and the best number of topics is 4. The method can obtain the satisfaction and dissatisfaction of tourists in a scenic spot in online evaluation data, thus avoiding an unpleasant feeling in the process of tourism. The scenic spot’s management efficiency can also be improved according to the situation.

https://doi.org/10.5573/IEIESPC.2024.13.4.372

(Ruixue Zhang)

English reading is crucial for college and university English education. Hence, new methods, such as adaptive recommendations of reading resources, are needed to improve teaching quality. The study establishes a recommendation model that classifies resources and students in both directions. The ID3 algorithm was improved by introducing correlation coefficients to enhance the information gain formula. An English training institution training course was selected as the subject. The performance test results show that the performance of the recommendation model based on the improved ID3 algorithm is better than the standard ID3 algorithm and the traditional recommendation model performance. The average score of students improved to 18.01±1.07, and the satisfaction rate of students was 91.22%. The model evaluated using four indicators revealed improved performance. The results show that students with strong reading ability are more sensitive to the difficulties of reading resources, while students with relatively weak ability should focus on the topic and type of content.

https://doi.org/10.5573/IEIESPC.2024.13.4.383

(Deng Liu)

Smart tourism is of great significance in society, and its core is how to obtain and utilize tourism-related information efficiently to provide a better tourism experience. This paper proposes a data mining method based on the Apriori association rule algorithm to solve the difficult search problem for complex and diverse tourism information. During the process, operator data are used as the data source for data mining, and the Apriori association rule algorithm is used as the foundation to construct a smart tourism information search method. The method is constrained by the tourists’ travel order data at different tourist locations, and multithreaded parallel computing of the data is achieved through a parallel computing framework. The experimental results show that the initial accuracy of the proposed method in mining data types can reach up to 97.8%. When testing the number of association rules, the proposed method only had 2317 association rules with a support level of 0.032. The proposed method had a runtime of only 13.6Ks when involving 50M data pieces in large-scale datasets, which was lower than other methods. Hence, the proposed method can effectively search for smart tourism information and has high search efficiency and data accuracy.

https://doi.org/10.5573/IEIESPC.2024.13.4.393

(Peng Zhang) ; (Wenjing Han) ; (Quanzhi Liu)

The mental health of college students is facing challenges because of the rapid changes in society. Anticipating these changes to enhance the emotional well-being of college students is crucial. This study devised a questionnaire focusing on pressure sources, such as employment and academic pressures. The mental health of college students was assessed using the SCL-90 scale, and data were collected as samples. A predictive model based on a back-propagation neural network (BPNN) was then constructed. The BPNN parameters were fine-tuned using the improved seagull optimization algorithm (ISOA), resulting in the ISOA-BPNN prediction model. The ISOA algorithm improved the BPNN prediction performance significantly compared to optimization algorithms, such as particle swarm optimization (PSO) and artificial bee colony (ABC), achieving an accuracy of 0.9762, an F1 value of 0.9834, and an area under the curve (AUC) of 0.9956. The ISOA-BPNN model demonstrated superior performance in predicting the mental health status of college students compared to prediction models, such as Logistic regression. These findings confirm the reliability of the ISOA-BPNN model developed in this study for predicting the mental health of college students and its potential applicability.

https://doi.org/10.5573/IEIESPC.2024.13.4.402

(Penglin Yu)

The modern logistics industry is facing significant data security challenges. The decentralized and transparent nature of blockchain technology provides a solution strategy. This study designed a logistics system based on the blockchain. In addition, a data security guarantee system was constructed, and the performance was tested. This study used a probability output regression algorithm. The system delay increased from 220 ms to 420 ms when the number of nodes was increased to 110. On the other hand, the PoR algorithm that introduces the reputation mechanism exhibited a lower initial delay (143). Moreover, when the number of nodes was increased to 110, its delay decreased by 90 ms compared to the PoR. When the number of nodes was increased to 60, the throughput of the PoR exceeded that of the reputation mechanism. The performance test of logistics data sharing showed that the processing time dimension was data-intensive, with a distribution density of 0.68. During the data-sharing process, 2350ms is required for every thousand encryptions; 3154 ms and 4263 ms were required for re-encryption conversion and decryption, respectively. This system performed well in terms of performance and data security, effectively handling large-scale logistics data, reducing the system latency, improving throughput, and ensuring data security. This result has profound significance in promoting the digital transformation of the logistics industry, improving operational efficiency and service quality.

https://doi.org/10.5573/IEIESPC.2024.13.4.414

(Yanfei Gao)

With the rapid development of smart communities, the volume of data in the community policing security system is increasing, which poses new challenges to behavioral anomaly monitoring. This paper proposes a behavioral anomaly monitoring method based on the One-Class Support Vector Machine (OC-SVM) algorithm and the MobileNetV3 improved OpenPose model. The fundamental theories of human posture recognition and behavioral anomaly detection are first reviewed, followed by a discussion of the limitations of the traditional OpenPose model. The structure of the MobileNetV3 lightweight network and its application advantages in human posture detection are described in detail. In this study, the feature extraction process was optimized using parallel computing technology to address the shortcomings of the OpenPose model, and an efficient method for monitoring behavioral anomalies was proposed. The test results of the system on the Multiple Cameras Fall dataset showed that the proposed method achieved 86.9% and 84.3% precision and recall for fall detection, respectively, showing significantly improved detection performance compared to the traditional method. The algorithm improvement in this paper provides a new solution for police security monitoring in smart communities, ensuring the safety and well-being of community residents.