• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.6
• /
• pp.319-327
• /
• 2022

Transformers are the most famous deep learning models that has achieved great success in natural language processing and also showed good performance on computer vision. In this survey, we categorized transformer-based models for computer vision, particularly object detection tasks and perform comprehensive comparative experiments to understand the characteristics of each model. Next, we evaluated the models subdivided into standard transformer, with key point attention, and adding attention with coordinates by performance comparison in terms of object detection accuracy and real-time performance. For performance comparison, we used two metrics: frame per second (FPS) and mean average precision (mAP). Finally, we confirmed the trends and relationships related to the detection and real-time performance of objects in several transformer models using various experiments.

• Journal of the Korea Society for Simulation
• /
• v.31 no.4
• /
• pp.11-22
• /
• 2022

Unlike a short-range, a long-range underwater acoustic communication(UWAC) uses low frequency signal and deep sound channel to minimize propagation loss. In this case, even though communication signals are modulated using a covert transmission technique such as spread spectrum, it is hard to conceal the existence of the signals. The unconcealed communication signal can be utilized as active sonar signal by enemy and presence of underwater vehicles may be exposed to the interceptor. Since it is very important to maintain stealthiness for underwater vehicles, the detection probability of friendly underwater vehicles should be considered when interceptor utilizes our long-range UWAC signal. In this paper, we modeled a long-range UWAC environment for analyzing the detection performance of underwater vehicles and proposed the region of interest(ROI) setup method and the measurement of detection performance. By computer simulations, we yielded parameters, analyzed the detection probability and the detection performance in ROI. The analysis results showed that the proposed detection performance analysis method for underwater vehicles could play an important role in the operation of long-range UWAC equipment.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.40 no.9
• /
• pp.1722-1724
• /
• 2015

This paper proposes a partial detection scheme using QRD-M, DFE, and iterative schemes for efficiency in terms of detection performance and complexity in a MIMO-OFDM system. The proposed scheme detects signals by using the different detection methods in according to spatial stream. In the proposed scheme, QRD-M with high detection performance and high complexity is used in spatial stream that requires low complexity, and DFE with low detection performance and low complexity is used in spatial stream that requires high complexity. Also, the iterative detection is performed in the detected spatial stream by using DFE. From the simulation, it is confirmed that although proposed scheme has increased complexity, detection performance is greatly improved by the proposed scheme.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.15 no.5
• /
• pp.243-249
• /
• 2020

Event detection in wireless sensor networks is a key requirement in many applications. Acoustic sensors are one of the most frequently used sensors for event detection in sensor networks, but they are sensitive and difficult to handle because they vary greatly depending on the environment and target characteristics of the sensor field. In this paper, we propose a learning-based improvement of CFAR algorithm for increasing node-level event detection performance in acoustic sensor networks, and verify the effectiveness of the designed algorithm by comparing and evaluating the event detection performance with other algorithms. Our experimental results demonstrate the superiority of the proposed algorithm by increasing the detection accuracy by more than 45.16% by significantly reducing false positives by 7.97 times while slightly increasing the false negative compared to the existing algorithm.

• International Journal of Advanced Culture Technology
• /
• v.11 no.1
• /
• pp.370-375
• /
• 2023

As we enter a super-aged society, studies are being conducted to reduce complications and deaths caused by falls in elderly adults. Research is being conducted on interventions for preventing falls in the elderly, wearable devices for detecting falls, and methods for improving the performance of fall detection algorithms. Wearable devices for detecting falls of the elderly generally use gyro sensors. In addition, to improve the performance of the fall detection algorithm, an artificial intelligence algorithm is applied to the x, y, z coordinate data collected from the gyro sensor. In this paper, we develop a wearable device that uses a gyro sensor, body temperature, and heart rate sensor for health management as well as fall detection for the elderly. In addition, we develop a fall detection and health management system that works with wearable devices and a guardian's mobile app to improve the performance of the fall detection algorithm and provide health information to guardians.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.8
• /
• pp.2688-2703
• /
• 2022

Research on the advanced detection of harmful objects in airport cargo for passenger safety against terrorism has increased recently. However, because associated studies are primarily focused on the detection of relatively large objects, research on the detection of small objects is lacking, and the detection performance for small objects has remained considerably low. Here, we verified the limitations of existing research on object detection and developed a new model called the Small Hazardous Object detection enhanced and reconstructed Model based on the You Only Look Once version 5 (YOLOv5) algorithm to overcome these limitations. We also examined the performance of the proposed model through different experiments based on YOLOv5, a recently launched object detection model. The detection performance of our model was found to be enhanced by 0.3 in terms of the mean average precision (mAP) index and 1.1 in terms of mAP (.5:.95) with respect to the YOLOv5 model. The proposed model is especially useful for the detection of small objects of different types in overlapping environments where objects of different sizes are densely packed. The contributions of the study are reconstructed layers for the Small Hazardous Object detection enhanced and reconstructed Model based on YOLOv5 and the non-requirement of data preprocessing for immediate industrial application without any performance degradation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.4
• /
• pp.1504-1526
• /
• 2018

For uplink multi-user massive MIMO systems, conventional minimum mean square error (MMSE) linear detection method achieves near-optimal performance when the number of antennas at base station is much larger than that of the single-antenna users. However, MMSE detection involves complicated matrix inversion, thus making it cumbersome to be implemented cost-effectively and rapidly. In this paper, we first summarize in detail the state-of-the-art simplified MMSE detection algorithms that circumvent the complicated matrix inversion and hence reduce the computation complexity from ${\mathcal{O}}(K^3)$ to ${\mathcal{O}}(K^2)$ or ${\mathcal{O}}(NK)$ with some certain performance sacrifice. Meanwhile, we divide the simplified algorithms into two categories, namely the matrix inversion approximation and the classical iterative linear equation solving methods, and make comparisons between them in terms of detection performance and computation complexity. In order to further optimize the detection performance of the existing detection algorithms, we propose more proper solutions to set the initial values and relaxation parameters, and present a new way of reconstructing the exact effective noise variance to accelerate the convergence speed. Analysis and simulation results verify that with the help of proper initial values and parameters, the simplified matrix inversion based detection algorithms can achieve detection performance quite close to that of the ideal matrix inversion based MMSE algorithm with only a small number of series expansions or iterations.

• Journal of Digital Convergence
• /
• v.15 no.6
• /
• pp.391-398
• /
• 2017

In this paper, we compared the performance of "Network Intrusion Detection System based on attack feature selection using fuzzy control language"[1] and "Intelligent Intrusion Detection System Model for attack classification using RNN"[2]. In this paper, we compare the intrusion detection performance of two techniques using KDD CUP 99 dataset. The KDD 99 dataset contains data sets for training and test data sets that can detect existing intrusions through training. There are also data that can test whether training data and the types of intrusions that are not present in the test data can be detected. We compared two papers showing good intrusion detection performance in training and test data. In the comparative paper, there is a lack of performance to detect intrusions that exist but have no existing intrusion detection capability. Among the attack types, DoS, Probe, and R2L have high detection rate using fuzzy and U2L has a high detection rate using RNN.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.13 no.3
• /
• pp.461-469
• /
• 2010

In this paper, we have proposed an improved fusion method of detection features which can enhance the detection probability under the given false alarm rate in the prescreening stage of SAR ATR(Synthetic Aperture Radar Automatic Target Recognition) system. Since the detection features have the positive correlation, the detection performance can be improved if the joint probability distribution of detection features is considered in the fusion process. The detection region is designed as a simple piecewise linear function which can be represented by few parameters. The parameters for the detection region can be derived by training the sample SAR images to maximize the detection probability with the given false alarm rate. Simulation result shows that the detection performance of the proposed method is improved for all combinations of detection features.

• International Journal of Internet, Broadcasting and Communication
• /
• v.11 no.4
• /
• pp.76-85
• /
• 2019

We introduce a vision-based object detection method for real-time video surveillance system in low-end edge computing environments. Recently, the accuracy of object detection has been improved due to the performance of approaches based on deep learning algorithm such as Region Convolutional Neural Network(R-CNN) which has two stage for inferencing. On the other hand, one stage detection algorithms such as single-shot detection (SSD) and you only look once (YOLO) have been developed at the expense of some accuracy and can be used for real-time systems. However, high-performance hardware such as General-Purpose computing on Graphics Processing Unit(GPGPU) is required to still achieve excellent object detection performance and speed. To address hardware requirement that is burdensome to low-end edge computing environments, We propose sub-frame analysis method for the object detection. In specific, We divide a whole image frame into smaller ones then inference them on Convolutional Neural Network (CNN) based image detection network, which is much faster than conventional network designed forfull frame image. We reduced its computationalrequirementsignificantly without losing throughput and object detection accuracy with the proposed method.