• The Journal of the Acoustical Society of Korea
• /
• v.40 no.3
• /
• pp.234-240
• /
• 2021

In this paper, the mask-based speech enhancement is improved for effective speech recognition in noise environments. In the mask-based speech enhancement, enhanced spectrum is obtained by multiplying the noisy speech spectrum by the mask. The VoiceFilter (VF) model is used as the mask estimation, and the Spectrogram Inpainting (SI) technique is used to remove residual noise of enhanced spectrum. In this paper, we propose a combined loss to further improve speech enhancement. In order to effectively remove the residual noise in the speech, the positive part of the Triplet loss is used with the component loss. For the experiment TIMIT database is re-constructed using NOISEX92 noise and background music samples with various Signal to Noise Ratio (SNR) conditions. Source to Distortion Ratio (SDR), Perceptual Evaluation of Speech Quality (PESQ), and Short-Time Objective Intelligibility (STOI) are used as the metrics of performance evaluation. When the VF was trained with the mean squared error and the SI model was trained with the combined loss, SDR, PESQ, and STOI were improved by 0.5, 0.06, and 0.002 respectively compared to the system trained only with the mean squared error.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.14 no.2
• /
• pp.134-139
• /
• 2021

Currently, the autonomous vehicle market is commercializing a level 3 autonomous vehicle, but it still requires the attention of the driver. After the level 3 autonomous driving, the most notable aspect of level 4 autonomous vehicles is vehicle stability. This is because, unlike Level 3, autonomous vehicles after level 4 must perform autonomous driving, including the driver's carelessness. Therefore, in this paper, we propose the Emergency-notification and Driver-response Confirmation System(EDCS) for an autonomousvehicle safety that notifies the driver of an emergency situation and recognizes the driver's reaction in a situation where the driver is careless. The EDCS uses the emergency situation delivery module to make the emergency situation to text and transmits it to the driver by voice, and the driver response confirmation module recognizes the driver's reaction to the emergency situation and gives the driver permission Decide whether to pass. As a result of the experiment, the HMM of the emergency delivery module learned speech at 25% faster than RNN and 42.86% faster than LSTM. The Tacotron2 of the driver's response confirmation module converted text to speech about 20ms faster than deep voice and 50ms faster than deep mind. Therefore, the emergency notification and driver response confirmation system can efficiently learn the neural network model and check the driver's response in real time.

• KIPS Transactions on Computer and Communication Systems
• /
• v.10 no.7
• /
• pp.191-198
• /
• 2021

Since the size of training dataset become large and the model is getting deeper to achieve high accuracy in deep learning, the deep neural network training requires a lot of computation and it takes too much time with a single node. Therefore, distributed deep learning is proposed to reduce the training time by distributing computation across multiple nodes. In this study, we propose hybrid allreduce strategy that considers the characteristics of each layer and communication and computational overlapping technique for synchronization of distributed deep learning. Since the convolution layer has fewer parameters than the fully-connected layer as well as it is located at the upper, only short overlapping time is allowed. Thus, butterfly allreduce is used to synchronize the convolution layer. On the other hand, fully-connecter layer is synchronized using ring all-reduce. The empirical experiment results on PyTorch with our proposed scheme shows that the proposed method reduced the training time by up to 33% compared to the baseline PyTorch.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.4
• /
• pp.611-620
• /
• 2021

The maintenance industry is mainly progressing based on condition-based maintenance after corrective maintenance and preventive maintenance. In condition-based maintenance, maintenance is performed at the optimum time based on the condition of equipment. In order to find the optimal maintenance point, it is important to accurately understand the condition of the equipment, especially the remaining useful life. Thus, using simulation data (C-MAPSS), a prediction model is proposed to predict the remaining useful life of a turbofan engine. For the modeling process, a C-MAPSS dataset was preprocessed, transformed, and predicted. Data pre-processing was performed through piecewise RUL, moving average filters, and standardization. The remaining useful life was predicted using principal component analysis and the k-NN method. In order to derive the optimal performance, the number of principal components and the number of neighbor data for the k-NN method were determined through 5-fold cross validation. The validity of the prediction results was analyzed through a scoring function while considering the usefulness of prior prediction and the incompatibility of post prediction. In addition, the usefulness of the RUL prediction model was proven through comparison with the prediction performance of other neural network-based algorithms.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.20 no.2
• /
• pp.95-111
• /
• 2021

As the population decreases in an aging society, the average age of drivers increases. Accordingly, the elderly at high risk of being in an accident need autonomous-driving vehicles. In order to secure driving safety on the road, several technologies to respond to various obstacles are required in those vehicles. Among them, technology is required to recognize static obstacles, such as poor road conditions, as well as dynamic obstacles, such as vehicles, bicycles, and people, that may be encountered while driving. In this study, we propose a deep neural network algorithm capable of simultaneously detecting these two types of obstacle. For this algorithm, we used 1,418 road images and produced annotation data that marks seven categories of dynamic obstacles and labels images to indicate road damage. As a result of training, dynamic obstacles were detected with an average accuracy of 46.22%, and road surface damage was detected with a mean intersection over union of 74.71%. In addition, the average elapsed time required to process a single image is 89ms, and this algorithm is suitable for personal mobility vehicles that are slower than ordinary vehicles. In the future, it is expected that driving safety with personal mobility vehicles will be improved by utilizing technology that detects road obstacles.

• Journal of the Korea Convergence Society
• /
• v.10 no.1
• /
• pp.103-113
• /
• 2019

License plate recognition camera is dedicated device designed for acquiring images of the target vehicle for recognizing letters and numbers in a license plate. Mostly, it is used as a part of the system combined with server and image analysis module rather than as a single use. However, building a system for vehicle license plate recognition is costly because it is required to construct a facility with a server providing the management and analysis of the captured images and an image analysis module providing the extraction of numbers and characters and recognition of the vehicle's plate. In this study, we would like to develop an embedded type convergent camera (Edge Base) which can expand the function of the camera to not only the license plate recognition but also the security CCTV function together and to perform two functions within the camera. This embedded type convergence camera equipped with a high resolution 4K IP camera for clear image acquisition and fast data transmission extracted license plate area by applying YOLO, a deep learning software for multi object recognition based on open source neural network algorithm and detected number and characters of the plate and verified the detection accuracy and recognition accuracy and confirmed that this camera can perform CCTV security function and vehicle number plate recognition function successfully.

• Journal of Korea Water Resources Association
• /
• v.51 no.12
• /
• pp.1207-1216
• /
• 2018

Discharge or water level predictions at tidally affected river reaches are currently still a great challenge in hydrological practices. This research aims to predict water level of the tide dominated site, Jamsu bridge in the Han River downstream. Physics-based hydrodynamic approaches are sometimes not applicable for water level prediction in such a tidal river due to uncertainty sources like rainfall forecasting data. In this study, TensorFlow deep learning framework was used to build a deep neural network based LSTM model and its applications. The LSTM model was trained based on 3 data sets having 10-min temporal resolution: Paldang dam release, Jamsu bridge water level, predicted tidal level for 6 years (2011~2016) and then predict the water level time series given the six lead times: 1, 3, 6, 9, 12, 24 hours. The optimal hyper-parameters of LSTM model were set up as follows: 6 hidden layers number, 0.01 learning rate, 3000 iterations. In addition, we changed the key parameter of LSTM model, sequence length, ranging from 1 to 6 hours to test its affect to prediction results. The LSTM model with the 1 hr sequence length led to the best performing prediction results for the all cases. In particular, it resulted in very accurate prediction: RMSE (0.065 cm) and NSE (0.99) for the 1 hr lead time prediction case. However, as the lead time became longer, the RMSE increased from 0.08 m (1 hr lead time) to 0.28 m (24 hrs lead time) and the NSE decreased from 0.99 (1 hr lead time) to 0.74 (24 hrs lead time), respectively.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.18 no.4
• /
• pp.58-70
• /
• 2019

In the case of a pedestrian traffic accident, it has a large-scale danger directly connected by a fatal accident at the time of the accident. The domestic ITS is not used for intelligent risk classification because it is used only for collecting traffic information despite of the construction of good quality traffic infrastructure. The CNN based pedestrian detection classification model, which is a major component of the proposed system, is implemented on an embedded system assuming that it is installed and operated in a restricted environment. A new model was created by improving YOLO's artificial neural network, and the real-time detection speed result of average accuracy 86.29% and 21.1 fps was shown with 20,000 iterative learning. And we constructed a protocol interworking scenario and implementation of a system that can connect with the ITS. If a pedestrian accident prevention system connected with ITS will be implemented through this study, it will help to reduce the cost of constructing a new infrastructure and reduce the incidence of traffic accidents for pedestrians, and we can also reduce the cost for system monitoring.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.29 no.3
• /
• pp.491-501
• /
• 2019

In side-channel analysis, which exploit physical leakage from a cryptographic device, deep learning based attack has been significantly interested in recent years. However, most of the state-of-the-art methods have been focused on classifying side-channel information in a profiled scenario where attackers can obtain label of training data. In this paper, we propose a new method based on deep learning to improve non-profiling side-channel attack such as Differential Power Analysis and Correlation Power Analysis. The proposed method is a signal preprocessing technique that reduces the noise in a trace by modifying Auto-Encoder framework to the context of side-channel analysis. Previous work on Denoising Auto-Encoder was trained through randomly added noise by an attacker. In this paper, the proposed model trains Auto-Encoder through the noise from real data using the noise-reduced-label. Also, the proposed method permits to perform non-profiled attack by training only a single neural network. We validate the performance of the noise reduction of the proposed method on real traces collected from ChipWhisperer board. We demonstrate that the proposed method outperforms classic preprocessing methods such as Principal Component Analysis and Linear Discriminant Analysis.

• Journal of the Korea Convergence Society
• /
• v.10 no.3
• /
• pp.1-6
• /
• 2019

In modern society, lifestyle and individuality are important, and personalized lifestyle and patterns are emerging. The number of people with articulation diseases is increasing due to wrong living habits. In addition, as the number of households increases, there is a case where emergency care is not received at the appropriate time. We need information that can be managed by ourselves through accurate analysis according to the individual's condition for health and disease management, and care appropriate to the emergency situation. It is effectively used for classification and prediction of data using CNN in deep learning. CNN differs in accuracy and processing time according to the data features. Therefore, it is necessary to improve processing speed and accuracy for real-time healthcare. In this paper, we propose motion monitoring using Mask R-CNN for articulation disease management. The proposed method uses Mask R-CNN which is superior in accuracy and processing time than CNN. After the user's motion is learned in the neural network, if the user's motion is different from the learned data, the control method can be fed back to the user, the emergency situation can be informed to the guardian, and appropriate methods can be taken according to the situation.