• IEMEK Journal of Embedded Systems and Applications
• /
• v.13 no.6
• /
• pp.289-296
• /
• 2018

Drone detection in FMCW radar system needs complex techniques because a drone beat frequency is highly dynamic and unpredictable. Therefore, the current static signal processing algorithms cannot show appropriate detection accuracy. With dynamic signal fluctuation and environmental clutters, it can fail to detect a drone or make false detection. It affects to the radar system integrity and safety. Constant false alarm rate (CFAR), one of famous static signal process algorithm is effective for static environment. But for drone detection, it shows low detection accuracy. In this paper, we suggest neural network based FMCW radar system for detecting a drone. We use recurrent neural network (RNN) because it is the effective neural network for signal processing. In our FMCW radar system, one transmitter emits FMCW signal and four-way fixed receivers detect reflected drone beat frequency. The coordinate of the drone can be calculated with four receivers information by triangulation. Therefore, RNN only learns and inferences reflected drone beat frequency. It helps higher learning and detection accuracy. With several drone flight experiments, RNN shows false detection rate and detection accuracy as 21.1% and 96.4%, respectively.

• Journal of Digital Convergence
• /
• v.18 no.2
• /
• pp.261-268
• /
• 2020

Machine learning is achieved by making a cost function from a given digital signal data and optimizing the cost function. The cost function here has local minimums in the cost function depending on the amount of digital signal data and the structure of the neural network. These local minimums make a problem that prevents learning. Among the many ways of solving these methods, our proposed method is to change the learning step-size. Unlike existed methods using the learning rate (step-size) as a fixed constant, the use of multivariate function as the cost function prevent unnecessary machine learning and find the best way to the minimum value. Numerical experiments show that the results of the proposed method improve about 3%(88.8%→91.5%) performance using the proposed method rather than the existed methods.

• The Journal of the Acoustical Society of Korea
• /
• v.43 no.2
• /
• pp.225-233
• /
• 2024

Active sonar transmits sound waves to detect covertly maneuvering underwater objects and detects the signals reflected back from the target. However, in addition to the target's echo, the active sonar's received signal is mixed with seafloor, sea surface reverberation, biological noise, and other noise, making target recognition difficult. Conventional techniques for detecting signals above a threshold not only cause false detections or miss targets depending on the set threshold, but also have the problem of having to set an appropriate threshold for various underwater environments. To overcome this, research has been conducted on automatic calculation of threshold values through techniques such as Constant False Alarm Rate (CFAR) and application of advanced tracking filters and association techniques, but there are limitations in environments where a significant number of detections occur. As deep learning technology has recently developed, efforts have been made to apply it in the field of underwater target detection, but it is very difficult to acquire active sonar data for discriminator learning, so not only is the data rare, but there are only a very small number of targets and a relatively large number of non-targets. There are difficulties due to the imbalance of data. In this paper, the image of the energy distribution of the detection signal is used, and a classifier is learned in a way that takes into account the imbalance of the data to distinguish between targets and non-targets and added to the existing technique. Through the proposed technique, target misclassification was minimized and non-targets were eliminated, making target recognition easier for active sonar operators. And the effectiveness of the proposed technique was verified through sea experiment data obtained in the East Sea.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.8
• /
• pp.2070-2079
• /
• 1997

In this paper, we proposed the FNCAC (fuzzy neural call admission control) scheme of the ATM networks which used the benefits of fuzzy logic controller and the learning abilities of the neural network to solve the call admission control problems. The new call in ATM networks is connected if QoS(quality of service) of the current calls is not affected due to the connection of a new call. The neural network CAC(call admission control) system is predictable system because the neural network is able to learn by the input/output pattern. We applied the fuzzy inference on the learning rate and momentum constant for improving the learning speed of the fuzzy neural network. The excellence of the proposed algorithm was verified using measurement of learning numbers in the traditional neural network method and fuzzy neural network method by simulation. We found that the learning speed of the FNCAC based on the fuzzy learning rules is 5 times faster than that of the CAC method based on the traditional neural network theory.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.58 no.4
• /
• pp.832-839
• /
• 2009

In this paper, we introduce an design of fuzzy relation-based fuzzy neural networks with multi-output. Fuzzy relation-based fuzzy neural networks comprise the network structure generated by dividing the entire input space. The premise part of the fuzzy rules of the network reflects the relation of the division space for the entire input space and the consequent part of the fuzzy rules expresses three types of polynomial functions such as constant, linear, and modified quadratic. For the multi-output structure the neurons in the output layer were connected with connection weights. The learning of fuzzy neural networks is realized by adjusting connections of the neurons both in the consequent part of the fuzzy rules and in the output layer, and it follows a back-propagation algorithm. In addition, in order to optimize the network, the parameters of the network such as apexes of membership functions, learning rate and momentum coefficient are automatically optimized by using real-coded genetic algorithm. Two examples are included to evaluate the performance of the proposed network.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.62 no.5
• /
• pp.705-711
• /
• 2013

In this paper, we introduce an design of multi-output fuzzy neural networks based on Interval Type-2 fuzzy set. The proposed Interval Type-2 fuzzy set-based fuzzy neural networks with multi-output (IT2FS-based FNNm) comprise the network structure generated by dividing the input space individually. The premise part of the fuzzy rules of the network reflects the individuality of the division space for the entire input space and the consequent part of the fuzzy rules expresses three types of polynomial functions with interval sets such as constant, linear, and modified quadratic inference for pattern recognition. The learning of fuzzy neural networks is realized by adjusting connections of the neurons in the consequent part of the fuzzy rules, and it follows a back-propagation algorithm. In addition, in order to optimize the network, the parameters of the network such as apexes of membership functions, uncertainty factor, learning rate and momentum coefficient were automatically optimized by using real-coded genetic algorithm. The proposed model is evaluated with the use of numerical experimentation.

• Proceedings of the KIEE Conference
• /
• 2000.07c
• /
• pp.1880-1884
• /
• 2000

The discrete wavelet transform is utilized as preprocessing of Neural Network(NN) to identify aging state of internal partial discharge in transformer. The discrete traveler transform is used to produce wavelet coefficients which are used for Classification. The statistical parameters (maximum of wavelet coefficients, average value, dispersion, skewness, kurtosis) using the wavelet coefficients are input into an back-propagation neural network. The neurons whose weights have obtained through Result of Cross-Validation. The Neural Network learning stops either when the error rate achieves an appropriate minimum or when the learning time overcomes a constant value. The networks, after training, can decide if the test signal is Early Aging State or Last Aging State or normal state.

• International conference on construction engineering and project management
• /
• 2022.06a
• /
• pp.736-743
• /
• 2022

Construction workers experience a high rate of fatal incidents from mobile equipment in the industry. One of the major causes is the decline in the acoustic condition of workers due to the constant exposure to construction noise. Previous studies have proposed various ways in which audio sensing and machine learning techniques can be used to track equipment's movement on the construction site but not on the audibility of safety signals. This study develops a novel framework to help automate safety surveillance in the construction site. This is done by detecting the audio sound at a different signal-to-noise ratio of -10db, -5db, 0db, 5db, and 10db to notify the worker of imminent dangers of mobile equipment. The scope of this study is focused on developing a signal processing model to help improve the audible sense of mobile equipment for workers. This study includes three-phase: (a) collect audio data of construction equipment, (b) develop a novel audio-based machine learning model for automated detection of collision hazards to be integrated into intelligent hearing protection devices, and (c) conduct field experiments to investigate the system' efficiency and latency. The outcomes showed that the proposed model detects equipment correctly and can timely notify the workers of hazardous situations.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.53 no.4
• /
• pp.195-200
• /
• 2004

Many kinds of speed sensorless control system of induction motor had been developed. But it is difficult to implement at the real system because of complex algorithm and equations. This paper investigates a novel speed sensorless control of induction motor using neural networks. The proposed control strategy is based on neural networks using stator current and output of neural model based on state observer. The errors between the stator current and the output of neural model are back-propagated to adjust the rotor speed, so that adaptive state variable will coincide with the desired state variable. This algorithm may overcome several shortages of conventional model, such as integrator problems, small EMF at low speed and relatively large sensitivity of stator resistance variation. Also, this paper presents a newly developed optimal equation about the momentum constant and the learning rate. The proposed algorithms are verified through simulation.

• Journal of Korean Society for Geospatial Information Science
• /
• v.4 no.2 s.8
• /
• pp.137-147
• /
• 1996

In this paper, the performances between maximum likelihood classifier based on statistical classification and multilayer perceptrons based on neural network approaches were compared and evaluated Experimental results from both neural network method and statistical method are presented. In addition, the nature of two different approches are analyzed based on the experiments.