• Science of Emotion and Sensibility
• /
• v.11 no.1
• /
• pp.1-10
• /
• 2008

We developed the decision support system by using the fuzzy theory. This system designs harmonious color space according to the linguistic input. This input represents the atmosphere which the user want. If the linguistic input of adjective image scale is given in the developed system, the relation between the adjective and color is supposed as fuzzy relation. The color which match with the whole atmosphere of color space is selected. The search region of harmonious color decision is controlled by the knowledge on color harmony of Moon-Spencer. Harmonious color is selected by it.

• Journal of Internet Computing and Services
• /
• v.18 no.6
• /
• pp.25-34
• /
• 2017

Autoencoder is a type of deep learning method where input layer and output layer are the same, and effectively extracts and restores characteristics of input vector using constraints of hidden layer. In this paper, we propose methods of Autoencoders to remove a natural background image which is a noise to the CAPTCHA and recover only a numerical images by applying various autoencoder models to a region where one number of CAPTCHA images and a natural background are mixed. The suitability of the reconstructed image is verified by using the softmax function with the output of the autoencoder as an input. And also, we compared the proposed methods with the other method and showed that our methods are superior than others.

• Proceedings of the KIEE Conference
• /
• 2002.07d
• /
• pp.2818-2820
• /
• 2002

In this paper, we design neurofuzzy networks architecture by means of linear fuzzy inference. The proposed neurofuzzy networks are equivalent to linear fuzzy rules, and the structure of these networks is composed of two main substructures, namely premise part and consequence part. The premise part of neurofuzzy networks use fuzzy space partitioning in terms of all variables for considering correlation between input variables. The consequence part is networks constituted as first-order linear form. The consequence part of neurofuzzy networks in general structure(for instance ANFIS networks) consists of nodes with a function that is a linear combination of input variables. But that of the proposed neurofuzzy networks consists of not nodes but networks that are constructed by connection weight and itself correspond to a linear combination of input variables functionally. The connection weights in consequence part are learned by back-propagation algorithm. For the evaluation of proposed neurofuzzy networks. The experimental results include a well-known NASA dataset concerning software cost estimation.

• Proceedings of the IEEK Conference
• /
• summer
• /
• pp.199-203
• /
• 2004

In this paper, we only describe the digital block of two-channel 18-bit analog-to-digital (A/D) converter employing sigma-delta method and xl28 decimation. The device contains two fourth comb filters with 1-bit input from sigma­delta modulator. each followed by a digital half band FIR(Finite Impulse Response) filters. The external analog sigma-delta modulators are sampled at 6.144MHz and the digital words are output at 48kHz. The fourth-order comb filter has designed 3 types of ways for optimal power consumption and signal-to-noise ratio. The following 3 digital filters are designed with 12tap, 22tap and 116tap to meet the specification. These filters eliminate images of the base band audio signal that exist at multiples of the input sample rate. We also designed these filters with 8bit and 16bit filter coefficient to analysis signal-to-noise ratio and hardware complexity. It also included digital output interface block for I2S serial data protocol, test circuit and internal input vector generator. It is fabricated with 0.35um HYNIX standard CMOS cell library with 3.3V supply voltage and the chip size is 2000um by 2000um. The function and the performance have been verified using Verilog XL logic simulator and Matlab tool.

• Journal of the Korea Society of Computer and Information
• /
• v.3 no.4
• /
• pp.1-9
• /
• 1998

Temporal Databases store all of informations by time varying, so the temporal query processor has to process very large information. Therefore, we propose an efficient method of query processing by using the relevance checking algorithm of input query and view definition. The relevance checking algorithm of query investigates relevance between the input query of user about base relation and the execution tree of view definition stored in system catalog. And related input query with view definition have a process of the query translation to the execution tree of view. So temporal query processor is able to increase performance of query processor by reducing the number of tuple.

• ETRI Journal
• /
• v.41 no.2
• /
• pp.235-241
• /
• 2019

In this paper, we propose a rank-weighted reconstruction feature to improve the robustness of a feed-forward deep neural network (FFDNN)-based acoustic model. In the FFDNN-based acoustic model, an input feature is constructed by vectorizing a submatrix that is created by slicing the feature vectors of frames within a context window. In this type of feature construction, the appropriate context window size is important because it determines the amount of trivial or discriminative information, such as redundancy, or temporal context of the input features. However, we ascertained whether a single parameter is sufficiently able to control the quantity of information. Therefore, we investigated the input feature construction from the perspectives of rank and nullity, and proposed a rank-weighted reconstruction feature herein, that allows for the retention of speech information components and the reduction in trivial components. The proposed method was evaluated in the TIMIT phone recognition and Wall Street Journal (WSJ) domains. The proposed method reduced the phone error rate of the TIMIT domain from 18.4% to 18.0%, and the word error rate of the WSJ domain from 4.70% to 4.43%.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.10a
• /
• pp.149-152
• /
• 2016

In this paper, we propose an algorithm to track human body of input video from a single camera. The proposed method gets the difference image between gray image of input image and one of background image and also the difference image between hue image of input image and one of background image. Then we combine the results, splits foreground and background and detect human body objects. Then each object is numbered and is tracked. The proposed method tracks each object using the intensity and hue histogram of local area in objects. The proposed method is applied to video from a camera and tracked well the hided objects and the overlapped objects.

• Journal of Korean Society of Occupational and Environmental Hygiene
• /
• v.32 no.3
• /
• pp.231-241
• /
• 2022

Objectives: The purpose of this study is to evaluate the input status of exposure-related information in the working environment monitoring database (WEMD) and special health examination database (SHED) for the construction of a national exposure surveillance system. Methods: The industrial and process code input status of WEMD and SHED for 21 carcinogens from 2014 to 2016 was compared. Data from workers who performed both work environment monitoring and special health examinations in 2019 and 2020 were extracted and the actual status of input of industrial and process codes was analyzed. We also investigated the cause of input errors through a focus group interview with 12 data input specialists. Results: As a result of analyzing WMED and SHED for 21 carcinogens, the five-digit industrial code matching rate was low at 53.5% and the process code matching rate was 19% or less. Among the data that simultaneously conducted work environment monitoring and special health examination in 2019 and 2020, the process code matching rate was very low at 18.1% and 5.2%, respectively. The main causes of exposure-related data input errors were the difference between the WEMD and SHED process code input systems from 2020, the number of standard process and job codes being too large, and the inefficiency of the standard code search system. Conclusions: In order to use WEMD and SHED as a national surveillance system, it is necessary to simplify the number of standard code input codes and improve the search system efficiency.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.12
• /
• pp.4275-4291
• /
• 2021

With the unprecedented growth of textual information on the Internet, an efficient automatic summarization system has become an urgent need. Recently, the neural network models based on the encoder-decoder with an attention mechanism have demonstrated powerful capabilities in the sentence summarization task. However, for paragraphs or longer document summarization, these models fail to mine the core information in the input text, which leads to information loss and repetitions. In this paper, we propose an abstractive document summarization method by applying guidance signals of key sentences to the encoder based on the hierarchical encoder-decoder architecture, denoted as KI-HABS. Specifically, we first train an extractor to extract key sentences in the input document by the hierarchical bidirectional GRU. Then, we encode the key sentences to the key information representation in the sentence level. Finally, we adopt key information representation guided selective encoding strategies to filter source information, which establishes a connection between the key sentences and the document. We use the CNN/Daily Mail and Gigaword datasets to evaluate our model. The experimental results demonstrate that our method generates more informative and concise summaries, achieving better performance than the competitive models.

• Journal of Institute of Control, Robotics and Systems
• /
• v.20 no.2
• /
• pp.170-174
• /
• 2014

In this paper, an unknown input estimation method via the optimal FIR smoother is proposed for linear discrete-time systems. The unknown inputs are represented by random walk processes and treated as auxiliary states in augmented state space models. In order to estimate augmented states which include unknown inputs, the optimal FIR smoother is applied to the augmented state space model. Since the optimal FIR smoother is unbiased and independent of any a priori information of the augmented state, the estimates of each unknown input are independent of the initial state and of other unknown inputs. Moreover, the proposed method can be applied to stochastic singular systems, since the optimal FIR smoother is derived without the assumption that the system matrix is nonsingular. A numerical example is given to show the performance of the proposed estimation method.