• The Journal of the Convergence on Culture Technology
• /
• v.6 no.2
• /
• pp.509-514
• /
• 2020

The conventional TTS system consists of several modules, including text preprocessing, parsing analysis, grapheme-to-phoneme conversion, boundary analysis, prosody control, acoustic feature generation by acoustic model, and synthesized speech generation. But TTS system with deep learning is composed of Text2Mel process that generates spectrogram from text, and vocoder that synthesizes speech signals from spectrogram. In this paper, for the optimal Korean TTS system construction we apply Tacotron2 to Tex2Mel process, and as a vocoder we introduce the methods such as WaveNet, WaveRNN, and WaveGlow, and implement them to verify and compare their performance. Experimental results show that WaveNet has the highest MOS and the trained model is hundreds of megabytes in size, but the synthesis time is about 50 times the real time. WaveRNN shows MOS performance similar to that of WaveNet and the model size is several tens of megabytes, but this method also cannot be processed in real time. WaveGlow can handle real-time processing, but the model is several GB in size and MOS is the worst of the three vocoders. From the results of this study, the reference criteria for selecting the appropriate method according to the hardware environment in the field of applying the TTS system are presented in this paper.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.3
• /
• pp.33-43
• /
• 2022

In this paper, we propose an ensemble model facilitated by multi-channel palm images with attention U-Net models and pretrained convolutional neural networks (CNNs) for establishing a contactless palm-based user identification system using conventional inexpensive camera sensors. Attention U-Net models are used to extract the areas of interest including hands (i.e., with fingers), palms (i.e., without fingers) and palm lines, which are combined to generate three channels being ped into the ensemble classifier. Then, the proposed palm information-based user identification system predicts the class using the classifier ensemble with three outperforming pre-trained CNN models. The proposed model demonstrates that the proposed model could achieve the classification accuracy, precision, recall, F1-score of 98.60%, 98.61%, 98.61%, 98.61% respectively, which indicate that the proposed model is effective even though we are using very cheap and inexpensive image sensors. We believe that in this COVID-19 pandemic circumstances, the proposed palm-based contactless user identification system can be an alternative, with high safety and reliability, compared with currently overwhelming contact-based systems.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.6
• /
• pp.52-57
• /
• 1999

Funtional metal prototypes are often required in numerous industrial applications. These components are typically needed in the early stage of a project to determine form, fit and function. Recent R/P(Rapid Prototyping) part are made of soft materials such as plastics, wax, paper, these master models cannot be employed durable test in real harsh working environment. Parts by direct metal rapid tooling method, such as laser sintering, by now are hard to get net shape, pores of the green parts of powder casting method must be infiltrated to get proper strength as tool, and new type of 3D direct tooling system combining fabrication welding arc and cutting process is reported by song etc. But a system which can build directly 3D parts of high performance functional material as metal part would need long period of system development, massive investment and other serious obstacles, such as patent. In this paper, through the rapid tooling process as silicon rubber molding using R/P master model, and fabricate wax pattern in that silicon rubber mold using vacuum casting method, then we tranlsated the wax patterns to numerous metal prototypes by new investment casting process combined conventional investment casting with rapid pototyping & rapid tooling process. with this wax-injection-mold-free investment casting, we developed new investment casting process of fabricating numerous functional metal prototypes from one master model, combined 3-D CAD, R/P and conventional investment casting and tried to expect net shape measuring total dimension shrinkage from R/P part to metal part.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.5
• /
• pp.1117-1124
• /
• 2010

PMIPv6 is a network-based mobility support scheme, proposed and standardized by NetLMM WG of IETF. It is proposed to solve problems of conventional mobility schemes, and to improve inefficiency of those. The standard document describes network components and detailed procedures to provide mobility to MN. But it describes only a handover procedure between MAGs, not between LMAs. In order to support seamless connectivity of MN efficiently, a handover procedure between LMAs is necessary. The proposed scheme in this paper utilizes a route optimization procedure to prevent inefficiency of inter-LMA tunneling scheme. At the same time, the proposed scheme utilizes a pre-authentication scheme to reduce handover latency. According to the result of performance evaluations, the proposed scheme greatly reduces handover latency, compared to conventional mobility support schemes.

• International Journal of Computer Science & Network Security
• /
• v.23 no.9
• /
• pp.8-16
• /
• 2023

In Japan, high-speed ground transportation service using linear motors at speeds of 500 km/h is scheduled to begin in 2027. To accommodate 5G services in trains, a subcarrier spacing frequency of 30 kHz will be used instead of the typical 15 kHz subcarrier spacing to mitigate Doppler effects in such high-speed transport. Furthermore, to increase the cell size of the 5G mobile system, multiple base station antennas will transmit identical downlink (DL) signals to form an expanded cell size along the train rails. In this situation, the forward and backward antenna signals are Doppler-shifted in opposite directions, respectively, so the receiver in the train may suffer from estimating the exact Channel Transfer Function (CTF) for demodulation. In a previously published paper, we proposed a channel estimator based on Delay and Doppler Profiler (DDP) in a 5G SISO (Single Input Single Output) environment and successfully implemented it in a signal processing simulation system. In this paper, we extend it to 2×2 MIMO (Multiple Input Multiple Output) with spatial multiplexing environment and confirm that the delay and DDP based channel estimator is also effective in 2×2 MIMO environment. Its simulation performance is compared with that of a conventional time-domain linear interpolation estimator. The simulation results show that in a 2×2 MIMO environment, the conventional channel estimator can barely achieve QPSK modulation at speeds below 100 km/h and has poor CNR performance versus SISO. The performance degradation of CNR against DDP SISO is only 6dB to 7dB. And even under severe channel conditions such as 500km/h and 8-path inverse Doppler shift environment, the error rate can be reduced by combining the error with LDPC to reduce the error rate and improve the performance in 2×2 MIMO. QPSK modulation scheme in 2×2 MIMO can be used under severe channel conditions such as 500 km/h and 8-path inverse Doppler shift environment.

• Korean Journal of Ecology and Environment
• /
• v.38 no.4 s.114
• /
• pp.482-493
• /
• 2005

The objective of this research was to evaluate an sampling effect of fish species and individual number on sampling techniques of electrofishing and conventional capture methods of cast net and kick net in 38 stream sites sampled during June - September, 2005. For the study, sampling gears were categorized as three types of electrofishing method (EM), conventional sampling technique (CM), and the combined technique (CT) of the electrofishing and the conventional method to compare statistical differences. Major differences of species composition between the CM and EM method were found in some benthic species of Misgurnus mizolepis and Iksookimia koreensis along with lithophilic species of Pungtungia herzi. These species were predominated in the EM rather than the CM, indicating that conventional sampling can underestimate the abundance of benthic or lithophilic species. In contrast ,individual number of typical water column species such as Zacco platypus and Zacco temmincki were more sampled by the CM, so that these fish populations were community overestimated. Also, t-tests on the types of sampling gear from various watersheds of Chogang Stream, Yugu Stream, Daejeon Stream, and Gap Stream showed that total individual numbers and species number in each stream were significantly (t values = 2.806-6.896, p values<0.05, n = 5-14) greater in the CT than the CM. Similar statistical significance (p<1.001, n = (10-24) on sampling seasons were observed during the monsoon and postmonsoon. These results indicate that if the electrofishing is not added to the conventional gears, the abundance of fish population and community can be underestimated and some benthic or lithophilic species may be excluded from the analysis, resulting in overall errors including sampling, fish fauna, and final judgement of community abundance. Overall our results strongly suggest that new application of electrofishing method along with the conventional sampling gears reduce sampling bias on underestimation of the real fish populations and communities.

• Journal of People, Plants, and Environment
• /
• v.23 no.6
• /
• pp.655-665
• /
• 2020

Background and objective: This study was conducted to compare the cultivation environment, growth of cut flowers, and management performance of conventional farms and smart farms growing the standard cut chrysanthemum, 'Jinba'. Methods: Conventional and smart farms were selected, and facility information, cultivation environment, cut flower growth, and management performance were investigated. Results: The conventional and smart farms were located in Muan, Jeollanam-do, and conventional farming involved cultivating with soil culture in a plastic greenhouse, while the smart farm was cultivating with hydroponics in a plastic greenhouse. The conventional farm did not have sensors for environmental measurement such as light intensity and temperature and pH and EC sensors for fertigation, and all systems, including roof window, side window, thermal screen, and shading curtain, were operated manually. On the other hand, the smart farm was equipped with sensors for measuring the environment and nutrient solution, and was automatically controlled. The day and night mean temperatures, relative humidity, and solar radiation in the facilities of the conventional and the smart farm were managed similarly. But in the floral differentiation stage, the floral differentiation was delayed, as the night temperature of conventional farm was managed as low as 17.7℃ which was lower than smart farm. Accordingly, the harvest of cut flowers by the conventional farm was delayed to 35 days later than that of the smart farm. Also, soil moisture and EC of the conventional farm were unnecessarily kept higher than those of the smart farm in the early growth stage, and then were maintained relatively low during the period after floral differentiation, when a lot of water and nutrients were required. Therefore, growth of cut flower, cut flower length, number of leaves, flower diameter, and weight were poorer in the conventional farm than in the smart farm. In terms of management performance, yield and sales price were 10% and 38% higher for the smart farm than for the conventional farm, respectively. Also, the net income was 2,298 thousand won more for the smart farm than for the conventional farm. Conclusion: It was suggested that the improved growth of cut flowers and high management performance of the smart farm were due to precise environment management for growth by the automatic control and sensor.

• Asian-Australasian Journal of Animal Sciences
• /
• v.19 no.3
• /
• pp.363-367
• /
• 2006

The nutritional worth of Phalaris minor seeds was assessed in comparison to conventional cereal grains like maize and wheat. P. minor seeds had higher total ash and cell wall constituents as compared to wheat and maize grains. The CP content of P. minor was comparable to wheat grains but higher than maize grains. The in vitro studies revealed that the net gas production and availability of ME from P. minor was comparable to that of maize but the digestibility of nutrients was significantly (p<0.05) lower than that of conventional cereal grains. The digestion kinetic parameters for DM and CP revealed that P. minor had the highest (p<0.05) soluble fraction (a) followed by wheat and maize. Reverse trend was observed for insoluble but potentially degradable fraction (b). The effective and true DM and CP degradability was significantly (p<0.05) higher in wheat grains followed by that in P. minor and maize grains. The digestibility of OM and NDF was not affected by replacing cereal grains in concentrate mixture with P. minor seeds up to 75 per cent level. But the availability of ME from concentrate mixtures was comparable to control only up to 50% level of replacement. Replacement of cereal grains with P. minor did not affect the rapidly soluble fraction and insoluble but potentially degradable fraction of concentrate mixture containing P. minor up to 75 per cent, but it was depressed significantly at 100% replacement level. The effective and true degradability of DM of concentrate mixtures containing P. minor from 50 to 100 per cent was comparable to that of conventional concentrate mixture (CCM). The wheat based concentrate mixtures showed higher net gas production (208 vs. 201 ml/g DM/24 h), digestibility of nutrients and ME availability (9.64 vs. 9.54 MJ/kg DM) as compared to maize based concentrate mixture. The wheat based concentrate mixture had significantly (p<0.05) higher rumen undegradable fraction and effective degradability. The data conclusively revealed that conventional cereal grains could be replaced with P.minor seeds up to 75 per cent without affecting the availability of nutrients.

• Korean Journal of Ecology and Environment
• /
• v.53 no.1
• /
• pp.63-72
• /
• 2020

In this study, to discuss on the applicability of eDNA as a new method to investigate fish diversity at streams, we applied eDNA at 4 streams (Geum River, Ji Stream, Hwangji Stream, Seomjin River), where endangered species are inhabits, with conventional survey (cast net and kick net). The average (±standard deviation) number of species investigated by eDNA were 19 species (±4.4), and it was relatively higher than average of conventional survey, 10 species (±4.8). Most of case, in this study, eDNA was more efficient than conventional survey. However, there were errors on species identification of Korean endemic species and aliied species from eDNA, and it seems the universal primer (MiFish primer set) is not suitable for them. Furthermore, some of endangered species, caught by conventional method, was not detected by eDNA. As the present universal primer is not suitable for identify the every freshwater fish species in Korea, the complementing or development of universal primer is needed, and the eDNA application after species specific marker development for detecting specific species like endangered species should be considered. In conclusion, if the manual for field survey method by eDNA is developed, we expect applicability enlargement for water ecosystem survey.

• Bulletin of the Korean Chemical Society
• /
• v.24 no.3
• /
• pp.369-376
• /
• 2003

Coulson (ZINDO), Mulliken $(MP2/6-31G^*)$ and Natural $(MP2/6-31G^*)$ population analyses of several large molecules were performed by the Fragment Reassociation (FR) method. The agreement between the conventional ZINDO (or conventional MP2) and FR-ZINDO (or FR-MP2) charges of these molecules was excellent. The standard deviations of the FR-ZINDO net atomic charges from the conventional ZINDO net atomic charges were 0.0008 for $C_{10}H_{22}$ (32 atoms), 0.0012 for $NH_2-C_{16}O_2H_{28}-COOH$ (53 atoms), 0.0014 for $NH_3^+-C_{16}O_2H_{28}-COOH$ (54 atoms), 0.0017 for $NH_2-C_{16}O_2H_{28}-COO^-$ (52 atoms), 0.0019 for $NH_3^+-C_{16}O_2H_{28}-COO^-$ (53 atoms), 0.0024 for a conjugated model $(O=CH-(CH=CH)_{15}-C=O-(CH=CH)_{12}-CH=CH_2)$, 118 atoms), 0.0038 for aglycoristocetin $(C_{60}N_7O_{19}H_{52}^+$, 138 atoms), 0.0023 for a polypropylene model complexed with a zirconocene catalyst $(C_{68}H-{121}Zr^+$, 190 atoms) and 0.0013 for magainin $(C_{112}N_{29}O_{28}SH_{177}$, 347 atoms), respectively. The standard deviations of the FR-MP2 Mulliken (or Natural) partial atomic charges from the conventional ones were 0.0016 (or 0.0016) for $C_{10}H_{22}$, 0.0019 (or 0.0018) for $NH_2-C_{16}O_2H_{28}-COOH$ and 0.0033 (or 0.0023) for $NH_3^+-C_{16}O_2H_{28}-COO^-$, respectively. These errors were attributed to the shape of molecules, the choice of fragments and the degree of ionic characters of molecules as well as the choice of methods. The CPU time of aglycoristocetin, conjugated model, polypropylene model complexed with zirconocene and magainin computed by the FR-ZINDO method was respectively 2, 4, 6 and 21 times faster than that by the normal ZINDO method. The CPU time of $NH_2-C_{16}O_2H_{28}-COOH\;and\;NH_3^+-C_{16}O_2H_{28}-COO^-$ computed by the FR-MP2 method was, respectively, 6 and 20 times faster than that by the normal MP2 method. The largest molecule calculated by the FR-ZINDO method was B-DNA (766 atoms). These results will enable us to compute atomic charges of huge molecules near future.