• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.850-858
• /
• 2022

To deploy Gate Recurrent Units (GRU) on resource-constrained embedded devices, this paper presents a reconfigurable FPGA-based GRU accelerator that enables structured compression. Firstly, a dense GRU model is significantly reduced in size by hybrid quantization and structured top-k pruning. Secondly, the energy consumption on external memory access is greatly reduced by the proposed reuse computing pattern. Finally, the accelerator can handle a structured sparse model that benefits from the algorithm-hardware co-design workflows. Moreover, inference tasks can be flexibly performed using all functional dimensions, sequence length, and number of layers. Implemented on the Intel DE1-SoC FPGA, the proposed accelerator achieves 45.01 GOPs in a structured sparse GRU network without batching. Compared to the implementation of CPU and GPU, low-cost FPGA accelerator achieves 57 and 30x improvements in latency, 300 and 23.44x improvements in energy efficiency, respectively. Thus, the proposed accelerator is utilized as an early study of real-time embedded applications, demonstrating the potential for further development in the future.

• KNOM Review
• /
• v.23 no.2
• /
• pp.1-10
• /
• 2020

Network Function Virtualization (NFV) is a key technology of 5G networks that has the advantage of enabling building and operating networks flexibly. However, NFV can complicate network management because it creates numerous virtual resources that should be managed. In NFV environments, service function chaining (SFC) composed of virtual network functions (VNFs) is widely used to apply a series of network functions to traffic. Therefore, it is required to dynamically allocate the right amount of computing resources or instances to SFC for meeting service requirements. In this paper, we propose Deep Q-Networks (DQN)-based auto-scaling to operate the appropriate number of VNF instances in SFC. The proposed approach not only resizes the number of VNF instances in SFC composed of multi-tier architecture but also selects a tier to be scaled in response to dynamic traffic forwarding through SFC.

• Plant Journal
• /
• v.18 no.4
• /
• pp.66-72
• /
• 2023

In order to respond to the demand for flexible operation of thermal power generation, development of natural sliding pressure operation that minimizes throttle loss by opening the turbine control valve 100% and maximize power generation efficiency in conjunction with ESS in order to quickly respond to fluctuations in the system frequency is required. The logic development of natural sliding pressure operation with ESS was developed to modify the existing logic at the power plant's top-level control logic such as the unit master, the boiler master and the turbine master. Cooperative control algorithms that complement the advantages and disadvantages of ESS operation (quick response, limited capacity) and power plant operation (slow response, continuous operation) not only improve efficiency when applied to actual power plants, but also respond quickly and flexibly to load demands to ensure system stability.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.281-283
• /
• 2021

The representative companies mentioned in the recommendation service in the domestic OTT(Over-the-top media service) market are YouTube and Netflix. YouTube, through various methods, started personalized recommendations in earnest by introducing an algorithm to machine learning that records and uses users' viewing time from 2016. Netflix categorizes users by collecting information such as the user's selected video, viewing time zone, and video viewing device, and groups people with similar viewing patterns into the same group. It records and uses the information collected from the user and the tag information attached to the video. In this paper, we propose a method to improve video media recommendation by automatically generating metadata of video media that was written by hand.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.284-286
• /
• 2021

With the continuous development of the Internet and digital, platforms are emerging to store large amounts of media data and provide customized services to individuals through online. Companies that provide these services recommend movies that suit their personal tastes to promote media consumption. Each company is doing a lot of research on various algorithms to recommend media that users prefer. Movies are divided into genres such as action, melodrama, horror, and drama, and the film's audio (music, sound effect, voice) is an important production element that makes up the film. In this research, based on movie trailers, we extract audio for each genre, check the commonalities of audio for each genre, distinguish movie genres through supervised learning of artificial intelligence, and propose a utilization method for generating metadata in the future.

• The Journal of the Acoustical Society of Korea
• /
• v.26 no.6
• /
• pp.259-268
• /
• 2007

This paper contributes to an improvement of the statistical bandwidth extension(BWE) system based on Hidden Markov Model(HMM). First, the existing HMM training method for BWE, which is suggested originally by Jax, is analyzed in comparison with the general Baum-Welch training method. Next, based on this analysis, a new HMM-based BWE method is suggested which adopts the Baum-Welch re-estimation algorithm instead of the Jax's to train HMM model. Conclusionally speaking, the Baum-Welch re-estimation algorithm is a generalized form of the Jax's training method. It is flexible and adaptive in modeling the statistical characteristic of training data. Therefore, it generates a better model to the training data, which results in an enhanced BWE system. According to experimental results, the new method performs much better than the Jax's BWE systemin all cases. Under the given test conditions, the RMS log spectral distortion(LSD) scores were improved ranged from 0.31dB to 0.8dB, and 0.52dB in average.

• Journal of Internet Computing and Services
• /
• v.1 no.2
• /
• pp.29-37
• /
• 2000

Asynchronous transfer mode is flexible to support multimedia communication services using asynchronous time-sharing and statistical multimedia techniques to the existing data communication area, ATM ABR service controls network traffic using feedback information on the network congestion situation in order to guarantee the demanded service qualities and the available cell rates, In this paper we apply the control method using queue length prediction to the formation of feedback information for more efficient ABR traffic control. If backward node receive the longer delayed feedback information on the impending congestion, the switch can be already congested from the uncontrolled arriving traffic and the fluctuation of queue length can be inefficiently high in the continuing time intervals, The feedback control method proposed in this paper predicts the queue length in the switch using the slope of queue length prediction function and queue length changes in time-series, The predicted congestion information is backward to the node, NLMS and neural network are used as the predictive control functions, and they are compared from performance on the queue length prediction. Simulation results show the efficiency of the proposed method compared to the feedback control method without the prediction, Therefore, we conclude that the efficient congestion and stability of the queue length controls are possible using the prediction scheme that can resolve the problems caused from the longer delays of the feedback information.

• Journal of Satellite, Information and Communications
• /
• v.7 no.3
• /
• pp.16-21
• /
• 2012

Recently, an importance of culture industry has been emphasized through an increased income level, spare time and changed values of modern people. And demands of the performance, arts, exhibit are steadily being increased. However the stage equipment depends on foreign manufactures on account of the inactive domestic technical skills. Especially in the lighting direction part, it is essential to control the lighting source and detect the moving line of actors but it generally uses the manual control type and realization of actor's moving line regardless of existing IT-based technologies. Also the system operation of existing sensor-based tracking and detecting technologies depends on the main lighting source of the stage. Therefore, this paper proposed LED lighting control system using the variable FOV and multi sensor-based tracking algorithm, which are possible to efficiently track the stage actors and direct the stage lights. Also we demonstrated the practicality and possibility of realization through the integrated experiment of the proposed system and implementation of the salient hardware, software. Additionally, the usefulness of proposed system was demonstrated using performance simulations and actual measurements of implemented sensor output.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.3
• /
• pp.461-467
• /
• 2003

This paper proposes the method of mobility implementation for location searching in the intelligence wireless ATM networks that expand and apply standard broadband signaling capabilities, and analyze the performance based on the numerical algorithm. The existing B-ISDN UNI protocol stack demands the location search mechanism to determine the location of mobile terminal in the wireless ATM networks because it use single protocol through the fixed PTP interface or PTM interface that don't support terminal mobility. The proposed method make possible the dynamic mobility at a part of wireless access by minimizing the signaling load without a falling-off in system performance by using the intelligence network technology according to the expansion of ATM and B-ISDN signaling integration based on the fixed networks. We implemented the performance analysis by MFC modeling based on numerical algorithm, and realized the efficiency of expenses by carrying out the comparative signaling performance evaluation to measure the relative gains of location search service in the intelligence wireless ATM system. The obtained results have the flexibility to operate in the public B-ISDN network environment without a change of existing B-ISDN/ATM NNI signaling reference to support the wireless ATM access system, and can easily expand to correspond to terminal mobility and various multimedia services in the next broadband PCS.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37C no.11
• /
• pp.1138-1147
• /
• 2012

In this paper, an efficient power control based MF-TDMA resource allocation scheme is proposed for next generation military satellite communication systems. The proposed scheme has the flexibility is used to support heterogeneous terminals with differ in transmission capabilities. The method can be divided into two parts : burst size calculation and burst structure determination. At first, we estimate the link budget taken into account a dynamic satellite link state variation. Then, applicable ACM level and burst size is chosen. In burst structure determination phase, we reorganize the burst structure in time-frequency domain by controlling limited power, bandwidth, time resources. In particular, we compensate the power spectral density among different terminals to integrate them in same transponder, Furthermore, we increase the packing efficiency by controlling the ACM level of the burst in applicable power spectral density range. Simulation results show that the method increase the spectral efficiency and burst packing efficiency. In addition, slot allocation rejection ratio is successfully reduced.