• Geophysics and Geophysical Exploration
• /
• v.3 no.3
• /
• pp.83-87
• /
• 2000

The noise prediction filter using a local/remote reference was developed to obtain a high quality data from seismic surveys over the area where seismic transmission power is limited. The method used in the noise prediction filter is a 3-layer neural network whose algorithm is backpropagation. A NRF (Noise Reduction Factor) value of about 3.0 was obtained with appling training and test data to the trained noise prediction filter. However, the scaling technique generally used for minimizing EM noise from electric and electromagnetic data cannot reduce seismic noise, since the technique can allow only amplitude difference between two time series measured at the primary and reference sites.

• International Journal of Naval Architecture and Ocean Engineering
• /
• v.7 no.2
• /
• pp.426-434
• /
• 2015

In this paper we proposed a dada-aided Doppler estimation method for underwater acoustic communication. The training sequence is non-dedicate, hence it can be designed for Doppler estimation as well as channel equalization. We assume the channel has been equalized and consider only flat-fading channel. First, based on the training symbols the theoretical received sequence is composed. Next the least square principle is applied to build the objective function, which minimizes the error between the composed and the actual received signal. Then an iterative approach is applied to solve the least square problem. The proposed approach involves an outer loop and inner loop, which resolve the channel gain and Doppler coefficient, respectively. The theoretical performance bound, i.e. the Cramer-Rao Lower Bound (CRLB) of estimation is also derived. Computer simulations results show that the proposed algorithm achieves the CRLB in medium to high SNR cases.

• The Bulletin of The Korean Astronomical Society
• /
• v.37 no.2
• /
• pp.157.2-157.2
• /
• 2012

Direct-conversion receivers often suffer from a DC-offset that is a by-product of the direct conversion process to baseband. In general, a basic approach to reduce the DC-offset is to do simple average of the baseband signal and remove the DC by subtracting the average. However, this gives rise to a residual DC offset which degrades the performance when the receiver adopts the coding schemes with high coding rates such as 8-PSK. Therefore, more advanced methods should be additionally required for better performance. While the training sequences are basically designed to have good auto-correlation properties to facilitate the channel estimation, they may be not good for the simultaneous estimation of the channel response and the DC-offset. Also the DC offset compensation under a bad condition does not give good results due to the estimation error. Correspondingly, the proposed scheme employs the two important points. First, the training sequence codes are divided into two groups by MSE(Mean Squared Errors) for estimating the channel taps and then SNR calculated from each group is compared to predefined threshold to do fine DC-offset estimation. Next, ON/OFF module is applied for preventing performance degradation by large estimation error under severe channel conditions. The simulation results of the proposed scheme shows good performances compared to the existing algorithm. As a result, this scheme is surely applicable to the receiver design in many communications systems.

• The Journal of the Acoustical Society of Korea
• /
• v.37 no.6
• /
• pp.475-482
• /
• 2018

In this paper, we present a study on data augmentation methods for DNN (Deep Neural Network)-based audio tagging. In this system, an audio signal is converted into a mel-spectrogram and used as an input to the DNN for audio tagging. To cope with the problem associated with a small number of training data, we augment the training samples using time stretching, pitch shifting, dynamic range compression, and block mixing. In this paper, we derive optimal parameters and combinations for the augmentation methods through audio tagging simulations.

• Journal of Korea Multimedia Society
• /
• v.24 no.8
• /
• pp.1035-1043
• /
• 2021

The shoulder is less stable than other joints, making it easier to onset of various shoulder disorders. In addition, limited range of motion and pain in the shoulder due to shoulder disorders restricts daily life and social activities. The problem with exercise therapy can be reduced in exercise effect by causing boredom through simple repetition of motion, thus reducing the patient's willingness to participate. Therefore, this paper aims to provide a treatment method that can induce active participation of patients by developing devices capable of passive, active, and resistance exercise and serious game contents using them. Furthermore, sEMG was used to verify whether the rotational exercise in the horizontal and vertical using serious game contents helps the shoulder movement actually. The measured sEMG signal was classified as 5 phases according to the angle of rotation and calculated the mean integrated EMG. The mean integrated EMG for the experimental results was higher in all phases when rotational was performed compared to those when both horizontal and vertical rotational exercise remained initial posture, indicating an increase in muscle activity.

• Nuclear Engineering and Technology
• /
• v.54 no.2
• /
• pp.644-652
• /
• 2022

During the pyroprocessing operation, various signals can be collected by process monitoring (PM). These signals are utilized to diagnose process states. In this study, feasibility of using PM for nuclear safeguards of electrorefining operation was examined based on the use of machine learning for detecting off-normal operations. The off-normal operation, in this study, is defined as co-deposition of key elements through reduction on cathode. The monitored process signal selected for PM was cathode potential. The necessary data were produced through electrodeposition experiments in a laboratory molten salt system. Model-based cathodic surface area data were also generated and used to support model development. Computer models for classification were developed using a series of recurrent neural network architectures. The concept of transfer learning was also employed by combining pre-training and fine-tuning to minimize data requirement for training. The resulting models were found to classify the normal and the off-normal operation states with a 95% accuracy. With the availability of more process data, the approach is expected to have higher reliability.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.456-457
• /
• 2021

Recently, in order to more accurately provide indoor location-based services, technologies using Wi-Fi fingerprints and deep learning are being studied. Among the deep learning models, an RNN model that can store information from the past can store continuous movements in indoor positioning, thereby reducing positioning errors. When using an RNN model for indoor positioning, the collected training data must be continuous sequential data. However, the Wi-Fi fingerprint data collected to determine specific location information cannot be used as training data for an RNN model because only RSSI for a specific location is recorded. This paper proposes a region clustering technique for sequential input data generation of RNN models based on Wi-Fi fingerprint data.

• Journal of the Institute of Convergence Signal Processing
• /
• v.3 no.1
• /
• pp.8-14
• /
• 2002

In this paper, we realized a real-time text-independent speaker recognition system using gaussian mixture model, and applied frame level likelihood normalization method which shows its effects in verification system. The system has three parts as front-end, training, recognition. In front-end part, cepstral mean normalization and silence removal method were applied to consider speaker's speaking variations. In training, gaussian mixture model was used for speaker's acoustic feature modeling, and maximum likelihood estimation was used for GMM parameter optimization. In recognition, likelihood score was calculated with speaker models and test data at frame level. As test sentences, we used text-independent sentences. ETRI 445 and KLE 452 database were used for training and test, and cepstrum coefficient and regressive coefficient were used as feature parameters. The experiment results show that the frame-level likelihood method's recognition result is higher than conventional method's, independently the number of registered speakers.

• Journal of the Institute of Convergence Signal Processing
• /
• v.4 no.4
• /
• pp.58-65
• /
• 2003

This paper studies a steering control method using a neural network algorithm for an intelligent autonomous driving robot. Previous horizontal steering control methods were made by various possible situation on the road. However, it isn't possible to make out algorithms that consider all sudden variances on the road. In this paper, an intelligent steering control algorithm for an autonomous driving robot system is presented. The algorithm is based on Self Organizing Maps(SOM) and the feature points on the road are used as training datum. In a simulation test, it is available to handle a steering control using SOM for an autonomous steering control. The algorithm is evaluated on an autonomous driving robot. The algorithm is available to control a steering for an autonomous driving robot with better performance at the experiments.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.2 no.2
• /
• pp.95-99
• /
• 2002

The Loose Part Monitoring System(LPMS) has been designed to detect. locate and evaluate detached or loosened parts and foreign objects in the reactor coolant system. In this paper, at first, we presents an application of the back propagation neural network. At the preprocessing step, the moving window average filter is adopted to reject the reject the low frequency background noise components. And then, extracting the acoustic signature such as Starting point of impact signal. Rising time. Half period. and Global time, they are used as the inputs to neural network . Secondly, we applied the neural network algorithm to LPMS in order to estimate the mass of loose parts. We trained the impact test data of YGN3 using the backpropagation method. The input parameter for training is Rising clime. Half Period amplitude. The result shored that the neural network would be applied to LPMS. Also, applying the neural network to thin practical false alarm data during startup and impact test signal at nuclear power plant, the false alarms are reduced effectively.