|
|
|
题名
|
作者
|
年代
|
出处
|
被引量
|
| 1 | Long Short-Term Memory Recurrent Neural Network-Based Acoustic Model Using Connectionist Temporal Classification on a Large-Scale Training Corpus显示文摘A Long Short-Term Memory(LSTM) Recurrent Neural Network(RNN) has driven tremendous improvements on an acoustic model based on Gaussian Mixture Model(GMM). However, these models based on a hybrid method require a forced aligned Hidden Markov Model(HMM) state sequence obtained from the GMM-based acoustic model. Therefore, it requires a long computation time for training both the GMM-based acoustic model and a deep learning-based acoustic model. In order to solve this problem, an acoustic model using CTC algorithm is proposed. CTC algorithm does not require the GMM-based acoustic model because it does not use the forced aligned HMM state sequence. However, previous works on a LSTM RNN-based acoustic model using CTC used a small-scale training corpus. In this paper, the LSTM RNN-based acoustic model using CTC is trained on a large-scale training corpus and its performance is evaluated. The implemented acoustic model has a performance of 6.18% and 15.01% in terms of Word Error Rate(WER) for clean speech and noisy speech, respectively. This is similar to a performance of the acoustic model based on the hybrid method. | Donghyun Lee Minkyu Lim Hosung Park Yoseb Kang Jeong-Sik Park Gil-Jin Jang Ji-Hwan Kim | 2017 | China Communications2017，14，9： | 7 |
| 2 | Al-incorporation into Li7La3Zr2O12 solid electrolyte keeping stabilized cubic phase for all-solid-state Li batteries显示文摘We observe the influence of Al occupancies in Li sites on the formation process of the garnet solid electrolyte of Li_7La_3Zr_2O_(12)(LLZO).A direct incorporation of Al is first promoted in a Li-insufficient garnet solid electrolyte during the calcination process of 850°C and then the cubic phase of LLZO is obtained after successive annealing step of 1000°C.Comparing to pristine LLZO,Al incorporated LLZO shows less formation of Li_2CO_3,keeping crystallographic and physicochemical properties.This Al incorporation improves both the ionic conductivity and interfacial resistance to poisoning procedure. | Changbin Im Dongwon Park Hosung Kim Jaeyoung Lee | 2018 | Journal of Energy Chemistry2018，27，5： | 4 |
| 3 | Spatial patterns of water diffusion along white matter tracts in temporal lobe epilepsy显示文摘 | Luis Concha Hosung Kim Andrea Bernasconi Boris C. Bernhardt Neda Bernasconi | 2012 | Neurology2012，，5： | 1 |
| 4 | The Effect of Topography on Water Wetting and Micro/Nano Tribological Characteristics of Polymeric Surfaces显示文摘 | Eui-Sung Yoon Seung Ho Yang Hosung Kong Ki-Hwan Kim | 2003 | Tribology Letters2003，，2： | 1 |
| 5 | New stopping criteria for iterative decoding of LDPC codes in H-ARQ systems显示文摘 | Beomkyu Shin Sang-Hyo Kim Hosung Park | 2013 | International Journal of Communica- tion Systems2013，26，11： | 1 |
| 6 | Volume Estimation of Small Scale Debris Flows Based on Observations of Topographic Changes Using Airborne LiDAR DEMs显示文摘This paper describes a geographic information system(GIS)-based method for observing changes in topography caused by the initiation, transport, and deposition of debris flows using highresolution light detection and ranging(LiDAR) digital elevation models(DEMs) obtained before and after the debris flow events. The paper also describes a method for estimating the volume of debris flows using the differences between the LiDAR DEMs. The relative and absolute positioning accuracies of the LiDAR DEMs were evaluated using a real-time precise global navigation satellite system(GNSS) positioning method. In addition, longitudinal and cross-sectional profiles of the study area were constructed to determine the topographic changes caused by the debris flows. The volume of the debris flows was estimated based on the difference between the LiDAR DEMs. The accuracies of the relative and absolute positioning of the two LiDAR DEMs were determined to be ±10 cm and ±11 cm RMSE, respectively, which demonstrates the efficiency of the method for determining topographic changes at an scale equivalent to that of field investigations. Based on the topographic changes, the volume of the debris flows in the study area was estimated to be 3747 m3, which is comparable with the volume estimated based on the data from field investigations. | Hosung KIM Seung Woo LEE Chan-Young YUNE Gihong KIM | 2014 | Journal of Mountain Science2014，11，3： | 1 |
| 7 | A highly efficientPV system using a series connection of DC-DC converter outputwith a photovoltaic panel显示文摘 | KIM Hosung KIM Jonghyun MIN Byungduk | 2009 | Renewable Energy2009，34，11： | 1 |
| 8 | SNR Enhancement of OTDR Using Biorthogonal Codes and Generalized Inverses显示文摘 | LEE DUCKEY YOON HOSUNG KIM PILHAN | 2005 | IEEE photonics technology letters2005，17，1： | 1 |
| 9 | Limitation of PMD compensation due to polarization-dependent loss in high-speed optical transmission links显示文摘 | Kim Na Young Lee Duckey Yoon Hosung | 2002 | IEEE Photon Technol Lett2002，14，1： | 1 |
| 10 | TP-MobNet: A Two-pass Mobile Network for Low-complexity Classification of Acoustic Scene显示文摘Acoustic scene classification(ASC)is a method of recognizing and classifying environments that employ acoustic signals.Various ASC approaches based on deep learning have been developed,with convolutional neural networks(CNNs)proving to be the most reliable and commonly utilized in ASC systems due to their suitability for constructing lightweight models.When using ASC systems in the real world,model complexity and device robustness are essential considerations.In this paper,we propose a two-pass mobile network for low-complexity classification of the acoustic scene,named TP-MobNet.With inverse residuals and linear bottlenecks,TPMobNet is based on MobileNetV2,and following mobile blocks,coordinate attention and two-pass fusion approaches are utilized.The log-range dependencies and precise position information in feature maps can be trained via coordinate attention.By capturing more diverse feature resolutions at the network’s end sides,two-pass fusions can also train generalization.Also,the model size is reduced by applying weight quantization to the trained model.By adding weight quantization to the trained model,the model size is also lowered.The TAU Urban Acoustic Scenes 2020 Mobile development set was used for all of the experiments.It has been confirmed that the proposed model,with a model size of 219.6 kB,achieves an accuracy of 73.94%. | Soonshin Seo Junseok Oh Eunsoo Cho Hosung Park Gyujin Kim Ji-Hwan Kim | 2022 | Computers, Materials & Continua2022，，11： | 0 |
| 11 | Shape-tailored whispering gallery microcavity lasers designed by transformation optics显示文摘Semiconductor microdisk lasers have great potential as low-threshold,high-speed,and small-form-factor light sources required for photonic integrated circuits because of their high-Q factors associated with long-lived whispering gallery modes(WGMs).Despite these advantages,the rotational symmetry of the disk shape restricts practical applications of the photonic devices because of their isotropic emission,which lacks directionality in far-field emission and difficulty in free-space out coupling.To overcome this problem,deformation of the disk cavity has been mainly attempted.However,the approach cannot avoid significant Q degradation owing to the broken rotational symmetry.Here,we first report a deformed shape microcavity laser based on transformation optics,which exploits WGMs free from Q degradation.The deformed cavity laser was realized by a spatially varying distribution of deep-sub-wavelength-scale(60 nm diameter)nanoholes in an InGaAsP-based multi-quantum-well heterostructure.The lasing threshold of our laser is one-third of that of the same shaped homogeneous laser and quite similar to that of a homogeneous microdisk laser.The results mean that Q spoiling caused by the boundary shape deformation is recovered by spatially varying nanohole density distribution designed by transformation optics and effective medium approximation. | YONG-HOON LEE HONGHWI PARK INBO KIM SANG-JUN PARK SUNGHWAN RIM BYOUNG JUN PARK MOOHYUK KIM YUSHIN KIM MYUNG-KI KIM WON SEOK HAN HOSUNG KIM HONGSIK PARK MUHAN CHOI | 2023 | Photonics Research2023，11，9： | 0 |
| 12 | Language Model Using Differentiable Neural Computer Based on Forget Gate-Based Memory Deallocation显示文摘A differentiable neural computer(DNC)is analogous to the Von Neumann machine with a neural network controller that interacts with an external memory through an attention mechanism.Such DNC’s offer a generalized method for task-specific deep learning models and have demonstrated reliability with reasoning problems.In this study,we apply a DNC to a language model(LM)task.The LM task is one of the reasoning problems,because it can predict the next word using the previous word sequence.However,memory deallocation is a problem in DNCs as some information unrelated to the input sequence is not allocated and remains in the external memory,which degrades performance.Therefore,we propose a forget gatebased memory deallocation(FMD)method,which searches for the minimum value of elements in a forget gate-based retention vector.The forget gatebased retention vector indicates the retention degree of information stored in each external memory address.In experiments,we applied our proposed NTM architecture to LM tasks as a task-specific example and to rescoring for speech recognition as a general-purpose example.For LM tasks,we evaluated DNC using the Penn Treebank and enwik8 LM tasks.Although it does not yield SOTA results in LM tasks,the FMD method exhibits relatively improved performance compared with DNC in terms of bits-per-character.For the speech recognition rescoring tasks,FMD again showed a relative improvement using the LibriSpeech data in terms of word error rate. | Donghyun Lee Hosung Park Soonshin Seo Changmin Kim Hyunsoo Son Gyujin Kim Ji-Hwan Kim | 2021 | Computers, Materials & Continua2021，，7： | 0 |
您所在的机构暂未开通此部分的功能，请联系管理员