• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.441-446
• /
• 2020

In this paper, we propose the α-Feature Map Scaling (α-FMS) method which extends the FMS method that was designed to enhance the discriminative power of feature maps of deep neural networks in Speaker Verification (SV) systems. The FMS derives a scale vector from a feature map and then adds or multiplies them to the features, or sequentially apply both operations. However, the FMS method not only uses an identical scale vector for both addition and multiplication, but also has a limitation that it can only add a value between zero and one in case of addition. In this study, to overcome these limitations, we propose α-FMS to add a trainable parameter α to the feature map element-wise, and then multiply a scale vector. We compare the performance of the two methods: the one where α is a scalar, and the other where it is a vector. Both α-FMS methods are applied after each residual block of the deep neural network. The proposed system using the α-FMS methods are trained using the RawNet2 and tested using the VoxCeleb1 evaluation set. The result demonstrates an equal error rate of 2.47 % and 2.31 % for the two α-FMS methods respectively.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.19 no.5
• /
• pp.162-177
• /
• 2020

The perception of traffic environment based on various sensors in autonomous driving system has a direct relationship with driving safety. Recently, as the perception model based on deep neural network is used due to the development of machine learning/in-depth neural network technology, a the perception model training and high quality of a training dataset are required. However, there are several realistic difficulties to collect data on all situations that may occur in self-driving. The performance of the perception model may be deteriorated due to the difference between the overseas and domestic traffic environments, and data on bad weather where the sensors can not operate normally can not guarantee the qualitative part. Therefore, it is necessary to build a virtual road environment in the simulator rather than the actual road to collect the traning data. In this paper, a training dataset collection process is suggested by diversifying the weather, illumination, sensor position, type and counts of vehicles in the simulator environment that simulates the domestic road situation according to the domestic situation. In order to achieve better performance, the authors changed the domain of image to be closer to due diligence and diversified. And the performance evaluation was conducted on the test data collected in the actual road environment, and the performance was similar to that of the model learned only by the actual environmental data.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.12
• /
• pp.2783-2790
• /
• 2010

In this paper, we first review LDPC codes in general and a belief propagation algorithm that works in logarithm domain. LDPC codes, which is chosen 802.11n for wireless local access network(WLAN) standard, require a large number of computation due to large size of coded block and iteration. Therefore, we presented three kinds of low computational algorithms for LDPC codes. First, sequential decoding with partial group is proposed. It has the same H/W complexity, and fewer number of iterations are required with the same performance in comparison with conventional decoder algorithm. Secondly, we have apply early stop algorithm. This method reduces number of unnecessary iterations. Third, early detection method for reducing the computational complexity is proposed. Using a confidence criterion, some bit nodes and check node edges are detected early on during decoding. Through the simulation, we knew that the iteration number are reduced by half using subset algorithm and early stop algorithm is reduced more than one iteration and computational complexity of early detected method is about 30% offs in case of check node update, 94% offs in case of check node update compared to conventional scheme. The LDPC decoder have been implemented in Xilinx System Generator and targeted to a Xilinx Virtx5-xc5vlx155t FPGA. When three algorithms are used, amount of device is about 45% off and the decoding speed is about two times faster than convectional scheme.

• Journal of Navigation and Port Research
• /
• v.41 no.4
• /
• pp.189-194
• /
• 2017

The Moving Search Radar System (MSRS) monitors sea areas by moving along the coast. Since the radar is initially aligned to the front of the vehicle, it is important to know the changes in the heading azimuth of the vehicle to quickly acquire the target azimuth from the radar after the MSRS has moved. The heading azimuth can be obtained using the gyro compass, the GPS compass or the electronic compass. The electronic compass is suitable for MSRS requiring fast maneuverability due to its small volume, short stabilization time and low price. However, using a geomagnetic sensor may result in an error due to the surrounding magnetic field. Errors can make early automatic tracking of the satellites difficult and can reduce the radar detection accuracy. Therefore, this paper proposes a method to automatically compensate for the error reflecting the correction value on the radar obtained by comparing the reference azimuth calculated by solving the geodesic inverse problem using two coordinates between the radar and the geostationary satellite with the actually-directed azimuth angle of the satellite antenna. The feasibility and convenience of the proposed method were verified by applying it to the MSRS in the field.

• The Journal of the Convergence on Culture Technology
• /
• v.4 no.1
• /
• pp.337-351
• /
• 2018

This paper discusses 5G IOT, Augmented Reality, Cloud Computing, Big Data, Future Autonomous Driving Vehicle technology, and presents 5G utilization of Pyeongchang Winter Olympic Games and Jeju Smart City model. The reason is that 5G is the main artery of the 4th industry.5G is the fourth industrial aorta because 5G is the core infrastructure of the fourth industrial revolution. In order for the AI, autonomous vehicle, VR / AR, and Internet (IoT) era to take off, data must be transmitted several times faster and more securely than before. For example, if you send a stop signal to LTE, which is a communication technology, to a remote autonomous vehicle, it takes a hundredth of a second. It seems to be fairly fast, but if you run at 100km / h, you can not guarantee safety because the car moves 30cm until it stops. 5G is more than 20 gigabits per second (Gbps), about 40 times faster than current LTE. Theoretically, the vehicle can be set up within 1 cm. 5G not only connects 1 million Internet (IoT) devices within a radius of 1 kilometer, but also has a speed delay of less than 0.001 sec. Steve Mollenkov, chief executive officer of Qualcomm, the world's largest maker of smartphones, said, "5G is a key element and innovative technology that will connect the future." With 5G commercialization, there will be an economic effect of 12 trillion dollars in 2035 and 22 million new jobs We can expect to see the effect of creation.

• Journal of Navigation and Port Research
• /
• v.39 no.4
• /
• pp.293-297
• /
• 2015

Underwater acoustic communications has been limited use for military purposes in the past. However, the fields of underwater applications expend to detection, submarine and communication in recent. The excessive multipath encountered in underwater acoustic communication channel is creating inter symbol interference, which is limiting factor to achieve a high data rate and bit error rate performance. To improve the performance of a received signal in underwater communication, many researchers have been studied for channel coding scheme with excellent performance at low SNR. In this paper, we applied BCJR decoder based ( 2,1,7 ) convolution codes and to compensate for the distorted data induced by the multipath, we applying the turbo equalization method. Through the underwater experiment on the Gyeungcheun lake located in Mungyeng city, we confirmed that turbo equalization structure of BCJR has better performance than hard decision and soft decision of Viterbi decoding. We also confirmed that the error rate of decoder input is less than error rate of $10^{-1}$, all the data is decoded. We achieved sucess rate of 83% through the experiment.

• Ecology and Resilient Infrastructure
• /
• v.2 no.3
• /
• pp.216-223
• /
• 2015

The conservation project to protect an endangered fish, Gobiobotia naktongensis was executed against declining the gene diversity of the fish after the construction of Youngju Dam in a sand-bed stream (Naeseongcheon Stream). We tried to move the populations of G. naktongensis from submerged planned sites to alternative habitats, bred artificially and augmented the juveniles to optimal habitat, and monitored the results of the restoration implementation. No entity of G. naktongensis was confirmed at the planned submerged sites despite attempting to capture more than 8 times and eventually the movement to alternative habitat could not be implemented. About 40 individuals of G. naktongensis were captured in the Naeseongcheon Stream and a total of 5,000 individuals were artificially spawned up. The population of juvenile inherited the genetic diversity from the brood stock. The bred juveniles were discharged at the selected optimal site that had a habit condition what was similar to their natural habitat. The micro-dispersion around the discharging area was found at the early stages of the augmentation. The re-capturing rate of discharged juveniles was reduced as time passed. The discharged juveniles seemed to adapt to the natural environment of Naeseongcheon Stream. The observation of their high abdominal distension and excrement demonstrated that the juveniles fed successfully in the discharging area. Therefore, securement of genetic homogeneity and enhancement of restoration population of G. naktongensis in upstream and downstream of Youngju Dam site from the artificial seed producing seemed to be primarily successful and long-term monitoring and analysis of the effect was expected to be necessary.

• Applied Chemistry for Engineering
• /
• v.16 no.6
• /
• pp.768-771
• /
• 2005

The deposition of silicon nitride ($SiN_x$) thin films was carried out on $SiO_2/Si$ substrate at room temperature by reactive dc magnetron sputtering. The analysis of deposited $SiN_x$ films using x-ray photoelectron spectroscopy indicated that the composition of $SiN_x$ films was Si-rich. The deposited $SiN_x$ thin films were annealed by varying annealing temperature and time. X-ray diffraction (XRD) analysis was performed in order to examine the crystallization of Si in $SiN_x$ thin films. The optical and electrical properties of $SiN_x$ thin films were measured for the observation of Si nanodot. As a result, we observed the XRD peaks that might be the Si crystals. As the annealing time and annealing temperature increased, the photoluminescence intensity of $SiN_x$ films gradually increased. The capacitance-voltage characteristics of $SiN_x$ film measured before and after annealing indicated that the trap effect of electrons or holes occurred due to the existence Si nanodots in the $SiN_x$ thin films.

• Journal of The Korean Ophthalmological Society
• /
• v.59 no.12
• /
• pp.1122-1128
• /
• 2018

Purpose: To compare the measurements of the rebound tonometer (RT), Goldmann applanation tonometer (GAT) and noncontact tonometer (NCT) in patients who underwent penetrating keratoplasty (PKP), and to evaluate the reproducibility of the RT measurements. Methods: This study included 19 PKP eyes and 28 normal eyes. We compared the intraocular pressure (IOP) measurements of the GAT and NCT with the RT in both groups using Spearman's correlation analysis and the Wilcoxon signed-rank test. The IOP, as measured with an RT in each group, was assessed with respect to reproducibility using the intraclass correlation coefficient (ICC). Results: In normal eyes, there was no significant difference in the measurements obtained with the RT, GAT, and NCT (p > 0.050). In the patient group, the RT measurements were not significantly different from those of the GAT (p = 0.872), but they were significantly lower than those obtained with the NCT (p = 0.011). However, the RT measurements showed a relatively high correlation with those of the GAT and NCT (r = 0.770 and 0.879, respectively). The ICC of the RT was 0.986 for the PKP eye group and 0.961 for the normal eye group, both of which were highly reproducible. Conclusions: In PKP eyes, the measurements obtained with the RT showed a relatively high correlation with those of the GAT and NCT, and the repeatability of the RT measurements was high and similar to those for normal eyes. The RT can therefore be considered a useful method for measuring the IOP in PKP eyes.

• Korean Journal of Comparative Education
• /
• v.23 no.5
• /
• pp.19-52
• /
• 2013

This study describes and compares the systems of school internal and external evaluation in South Korea and Sweden. Since education environment over the world has been changing commonly toward 'improving quality of schools' in 'decentralization of education', the more stress goes to 'education evaluation'. In this respect, school evaluation becomes a vital means to monitor schools and improve their weaknesses, as well as to enhance their strengths by informing the public of the results of evaluation. The research explores the systems of South Korean and Swedish school evaluation carried out by internal and external inspectors in order to compare the differences of two countries through qualitative methods by analyzing documents and articles related to school evaluation of two countries. Even though both South Korean and Swedish education have a 'school evaluation system as a unit', the approaches to it vary in accordance with the contexts of each country. In this regard, 'convergence' of having the same education policy of 'school evaluation' and 'divergence' of different approaches toward evaluation systems in each country have been examined in this study.