• Journal of Digital Contents Society
• /
• v.18 no.1
• /
• pp.93-99
• /
• 2017

A hybrid approach using Long Short Term Memory (LSTM) Recurrent Neural Network (RNN) has showed great improvement in speech recognition accuracy. For training acoustic model based on hybrid approach, it requires forced alignment of HMM state sequence from Gaussian Mixture Model (GMM)-Hidden Markov Model (HMM). However, high computation time for training GMM-HMM is required. This paper proposes an end-to-end approach for LSTM RNN-based Korean speech recognition to improve learning speed. A Connectionist Temporal Classification (CTC) algorithm is proposed to implement this approach. The proposed method showed almost equal performance in recognition rate, while the learning speed is 1.27 times faster.

• Journal of the Semiconductor & Display Technology
• /
• v.20 no.4
• /
• pp.10-15
• /
• 2021

Cache bypassing schemes have been studied to remove unnecessary updating the data in cache blocks. Among them, a statistics-based cache bypassing method for asymmetric-access caches is one of the most efficient approach for non-voliatile memories and shows the lowest cache access latency. However, it is proposed under the condition of the normal cache system, so further study is required for the hybrid cache architecture. This paper proposes a novel cache bypassing scheme, called hybrid bypassing block selector. In the proposal, the new model is established considering the SRAM region and the non-volatile memory region separately. Based on the model, hybrid bypassing decision block is implemented. Experiments show that the hybrid bypassing decision block saves overall energy consumption by 21.5%.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.1
• /
• pp.1-10
• /
• 2010

Recently heterogeneous storage system such as hybrid hard disk drive (H-HDD) combining flash memory and magnetic disk is launched, according as the read performance of NAND flash memory is enhanced as similar to that of hard disk drive (HDD) and the power consumption of NAND flash memory is reduced less than that of HDD. However, the read and write operations of NAND flash memory are slower than those of rotational disk. Besides, serious overheads are incurred on CPU and main memory in the case that intensive write requests to flash memory are repeatedly occurred. In this paper, we propose the Least Frequently Used-Hot scheme that replaces the data blocks whose reference frequency of read operation is low and update frequency of write operation is high, and the data flushing scheme that rearranges the data blocks into the multi-zone of the rotation disk. Experimental results show that the execution time of the proposed method is 38% faster than those of conventional LRU and LFU block replacement schemes in I/O performance aspect and the proposed method increases the life span of Non-Volatile Cache 40% higher than those of conventional LRU, LFU, FIFO block replacement schemes.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2013.02a
• /
• pp.396-396
• /
• 2013

The resistive random access memory (ReRAM) has several advantages to apply next generation non-volatile memory device, because of fast switching time, long retentions, and large memory windows. The high mobility of monolayered graphene showed several possibilities for scale down and electrical property enhancement of memory device. In this study, the monolayered graphene grown by chemical vapor deposition was transferred to $SiO_2$ (100 nm)/Si substrate and glass by using PMMA coating method. For formation of metal-oxide nanoparticles, we used a chemical reaction between metal films and polyamic acid layer. The 50-nm thick BPDA-PDA polyamic acid layer was coated on the graphene layer. Through soft baking at $125^{\circ}C$ or 30 min, solvent in polyimide layer was removed. Then, 5-nm-thick indium layer was deposited by using thermal evaporator at room temperature. And then, the second polyimide layer was coated on the indium thin film. After remove solvent and open bottom graphene layer, the samples were annealed at $400^{\circ}C$ or 1 hr by using furnace in $N_2$ ambient. The average diameter and density of nanoparticle were depending on annealing temperature and times. During annealing process, the metal and oxygen ions combined to create $In_2O_3$ nanoparticle in the polyimide layer. The electrical properties of $In_2O_3$ nanoparticle ReRAM such as current-voltage curve, operation speed and retention discussed for applictions of transparent and flexible hybrid ReRAM device.

• Proceedings of the IEEK Conference
• /
• 2008.06a
• /
• pp.649-650
• /
• 2008

In order to resolve mechanical limit in HDD, recently, the hybrid hard disk combining HDD and a flash memory was launched. In this paper, we propose a simulator for hybrid hard disk which considers redirection, flushing and spin-down function to complement the difference between HDD and hybrid hard disk. The simulator was implemented in linux kernel 2.6.20 by modifying system calls related with file system. The experiment shows that the power consumption of hybrid hard disk is 47% smaller than that of hard disk in laptop PC.

• Journal of information and communication convergence engineering
• /
• v.19 no.4
• /
• pp.234-240
• /
• 2021

As the number of pet dog-related businesses is rising rapidly, there is an increasing need for reliable pet dog health information systems for casual pet owners, especially those caring for older dogs. Our goal is to implement a mobile pre-diagnosis system that can provide a first-hand pre-diagnosis and an appropriate coping strategy when the pet owner observes abnormal symptoms. Our previous attempt, which is based on the fuzzy C-means family in inference, performs well when only relevant symptoms are provided for the query, but this assumption is not realistic. Thus, in this paper, we propose a hybrid inference structure that combines fuzzy association memory and a double-layered fuzzy C-means algorithm to infer the probable disease with robustness, even when noisy symptoms are present in the query provided by the user. In the experiment, it is verified that our proposed system is more robust when noisy (irrelevant) input symptoms are provided and the inferred results (probable diseases) are more cohesive than those generated by the single-phase fuzzy C-means inference engine.

• Earthquakes and Structures
• /
• v.25 no.1
• /
• pp.15-26
• /
• 2023

It is of great importance to assess the residual displacement demand in the performance-based seismic design. In this paper, a hybrid deep learning model for predicting the residual displacement spectra under near-fault (NF) ground motions is proposed by combining the long short-term memory network (LSTM) and back-propagation (BP) network. The model is featured by its capacity of predicting the residual displacement spectrum under a given NF ground motion while considering the effects of structural parameters. To construct this model, 315 natural and artificial NF ground motions were employed to compute the residual displacement spectra through elastoplastic time history analysis considering different structural parameters. Based on the resulted dataset with a total of 9,450 samples, the proposed model was finally trained and tested. The results show that the proposed model has a satisfactory accuracy as well as a high efficiency in predicting residual displacement spectra under given NF ground motions while considering the impacts of structural parameters.

• Structural Engineering and Mechanics
• /
• v.42 no.6
• /
• pp.783-797
• /
• 2012

A new hybrid meta-heuristic optimization algorithm is presented for design of structures. The algorithm is based on the concepts of the charged system search (CSS) and the particle swarm optimization (PSO) algorithms. The CSS is inspired by the Coulomb and Gauss's laws of electrostatics in physics, the governing laws of motion from the Newtonian mechanics, and the PSO is based on the swarm intelligence and utilizes the information of the best fitness historically achieved by the particles (local best) and by the best among all the particles (global best). In the new hybrid algorithm, each agent is affected by local and global best positions stored in the charged memory considering the governing laws of electrical physics. Three different types of structures are optimized as the numerical examples with the new algorithm. Comparison of the results of the hybrid algorithm with those of other meta-heuristic algorithms proves the robustness of the new algorithm.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.43 no.9 s.351
• /
• pp.31-37
• /
• 2006

Low-density parity-check (LDPC) codes are recently emerged due to its excellent performance. The standard for European high definition satellite digital video broadcast, DVB-S2 has adopted LDPC codes as a channel coding scheme. This paper proposes a DVB-S2 LDPC decoder architecture using a hybrid parity check matrix which is efficient in hardware implementation for both decoders and encoders. The hybrid H-matrices are constructed so that both the semi-random technique and the partly parallel structure can be applied to design encoders and decoders. Using the hybrid H-matrix scheme, the architecture of LDPC decoder for DVB-S2 can be very practical and efficient. In addition, we show a new Variable Node processor Unit (VNU) architecture to reuse the VNU for various code rates and optimized block memory placement to reuse. We design a DVB-S2 LDPC decoder of code rate 1/2 usng the proposed architecture. We estimate the performance of the DVB-S2 LDPC decoder and compare it with other decoders.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 1996.04a
• /
• pp.130-135
• /
• 1996

This paper presents a proof-concept investigation on the active vibration control of two hybrid smart structures (HSSs). The first one is consisting of a piezoelectric film (PF) actuator and an electro-rheological fluid(ERF) actuator, and the other is featured by a piezoceramic (PZT) actuator and a shape memory alloy (SMA) actuator. For the PF/ERF hybrid smart structure, both the increment of the damping ratios and the suppression of the tip deflections are evaluated in order to demonstrate control effectiveness of the PF actuator and ERF actuator and the hybrid actuation. For the PZT/SMA hybrid smart structure, the PZT actuator takes account of the high frequency excitation, while the SMA actuator exerts large vibration control force. The experimental results exhibit superior abilities of the hybrid actuation systems to tailor elastodynamic responses of the HSS rather than a single class of actuation system alone.