• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.34 no.3
• /
• pp.439-449
• /
• 2024

Recent rapid advancements in voice generation technology have enabled the natural synthesis of voices using text alone. However, this progress has led to an increase in malicious activities, such as voice phishing (voishing), where generated voices are exploited for criminal purposes. Numerous models have been developed to detect the presence of synthesized voices, typically by extracting features from the voice and using these features to determine the likelihood of voice generation.This paper proposes a new model for extracting voice features to address misuse cases arising from generated voices. It utilizes a deep learning-based audio codec model and the pre-trained natural language processing model BERT to extract novel voice features. To assess the suitability of the proposed voice feature extraction model for voice detection, four generated voice detection models were created using the extracted features, and performance evaluations were conducted. For performance comparison, three voice detection models based on Deepfeature proposed in previous studies were evaluated against other models in terms of accuracy and EER. The model proposed in this paper achieved an accuracy of 88.08%and a low EER of 11.79%, outperforming the existing models. These results confirm that the voice feature extraction method introduced in this paper can be an effective tool for distinguishing between generated and real voices.

• Resources Recycling
• /
• v.33 no.3
• /
• pp.3-11
• /
• 2024

This study addressed the cost structure of metallurgical plants for vanadium recovery or production, which were previously planned or implemented. Vanadium metallurgy consists of several sub-processes such as such as pretreatment, roasting, leaching, precipitation, and filtration, in order to finally produce vanadium pentoxide. Here, lots of costs should be spent for such plants, in which these costs are largely divided into CAPEX (Capital Expenditure) and OPEX (Operational Expenditure). As a result, the capacities (feed input rates) and vanadium contents are various along the target projects for this study. However, final production rates and grades of vanadium pentoxide showed relatively small differences. In addition, a noticeable correlation is found between capacities and specific operating costs, in that a steadily decreasing trend is described with a non-linear curve with around -0.3 power. Therefore, for the plant capacity below 100,000 tons per year, the specific operating cost rapidly decreases as the capacity increases, whereas the cost remains relatively stable in the range of 0.6 to 1.2 million tons per year of the capacity. From a technical perspective, effective optimization of the metallurgical process plant can be achieved by improving vanadium recovery rate in the pre-treatment and/or roasting-leaching processes. Finally, the results of this study should be updated through future research with on-going field verification and further detailed cost analysis.

• Journal of KIISE:Computer Systems and Theory
• /
• v.27 no.7
• /
• pp.619-628
• /
• 2000

Abstract Perspective views provide a powerful depth cue and thus aid the interpretation of complicated images. The main drawback of current perspective volume rendering is the long execution time. In this paper, we present an efficient perspective volume rendering algorithm based on coherency between rays. Two sets of templates are built for the rays cast from horizontal and vertical scanlines in the intermediate image which is parallel to one of volume faces. Each sample along a ray is calculated by interpolating neighboring voxels with the pre-computed weights in the templates. We also solve the problem of uneven sampling rate due to perspective ray divergence by building more templates for the regions far away from a viewpoint. Since our algorithm operates in object-order, it can avoid redundant access to each voxel and exploit spatial data coherency by using run-length encoded volume. Experimental results show that the use of templates and the object-order processing with run-length encoded volume provide speedups, compared to the other approaches. Additionally, the image quality of our algorithm improves by solving uneven sampling rate due to perspective ray di vergence.

• Proceedings of the Korean Society of Medical Physics Conference
• /
• 2003.09a
• /
• pp.78-78
• /
• 2003

Purpose: This event related fMRI study was to further our understanding about how different brain regions could contribute to effective access of specific information stored in long term memory. This experiment has allowed us to determine the brain regions involved in recognition of familiar faces among non familiar faces. Materials and Methods: Twelve right handed normal, healthy volunteer adults participated in face recognition experiment. The paradigm consists of two 40 familiar faces, 40 unfamiliar faces and control base with scrambled faces in a randomized order, with null events. Volunteers were instructed to press on one of two possible buttons of a response box to indicate whether a face was familiar or not. Incorrect answers were ignored. A 1.5T MRI system(GMENS) was employed to evaluate brain activity by using blood oxygen level dependent (BOLD) contrast. Gradient Echo EPI sequence with TR/TE= 2250/40 msec was used for 17 contiguous axial slices of 7mm thickness, covering the whole brain volume (240mm Field of view, 64 ${\times}$ 64 in plane resolution). The acquired data were applied to SPM99 for the processing such as realignment, normalization, smoothing, statistical ANOVA and statistical preference. Results/Disscusion: The comparison of familiar faces vs unfamiliar faces yielded significant activations in the medial temporal regions, the occipito temporal regions and in frontal regions. These results suggest that when volunteers are asked to recognize familiar faces among unfamiliar faces they tend to activate several regions frequently involved in face perception. The medial temporal regions are also activated for familiar and unfamiliar faces. This interesting result suggests a contribution of this structure in the attempt to match perceived faces with pre existing semantic representations stored in long term memory.

• Journal of KIISE:Software and Applications
• /
• v.31 no.9
• /
• pp.1233-1245
• /
• 2004

Though biometrics to authenticate a person is a good tool in terms of security and convenience, typical authentication algorithms using biometrics may not be executed on resource-constrained devices such as smart cards. Thus, to execute biometric processing on resource-constrained devices, it is desirable to develop lightweight authentication algorithm that requires only small amount of memory and computation. Also, among biological features, face is one of the most acceptable biometrics, because humans use it in their visual interactions and acquiring face images is non-intrusive. We present a new face authentication algorithm in this paper. Our achievement is two-fold. One is to present a face authentication algorithm with low memory requirement, which uses support vector machines (SVM) with the feature set extracted by genetic algorithms (GA). The other contribution is to suggest a method to reduce further, if needed, the amount of memory required in the authentication at the expense of verification rate by changing a controllable system parameter for a feature set size. Given a pre-defined amount of memory, this capability is quite effective to mount our algorithm on memory-constrained devices. The experimental results on various databases show that our face authentication algorithm with SVM whose input vectors consist of discriminating features extracted by GA has much better performance than the algorithm without feature selection process by GA has, in terms of accuracy and memory requirement. Experiment also shows that the number of the feature ttl be selected is controllable by a system parameter.

• Korean Chemical Engineering Research
• /
• v.58 no.4
• /
• pp.588-595
• /
• 2020

Prediction and control of nitrogen oxides (NO_x) emission is of great interest in industry due to stricter environmental regulations. Herein, we propose an artificial intelligence (AI)-based framework for prediction of NO_x emission. The framework includes pre-processing of data for training of neural networks and evaluation of the AI-based models. In this work, Long-Short-Term Memory (LSTM), one of the recurrent neural networks, was adopted to reflect the time series characteristics of NO_x emissions. A decision tree was used to determine a time window of LSTM prior to training of the network. The neural network was trained with operational data from a heating furnace. The optimal model was obtained by optimizing hyper-parameters. The LSTM model provided a reliable prediction of NO_x emission for both training and test data, showing an accuracy of 93% or more. The application of the proposed AI-based framework will provide new opportunities for predicting the emission of various air pollutants with time series characteristics.

• Progress in Superconductivity
• /
• v.12 no.2
• /
• pp.88-92
• /
• 2011

[ $MgB_2$ ]superconducting sheets have been fabricated using powder roll compaction method. Sheet-type $MgB_2$ bulk samples were successfully fabricated using the pre-reacted $MgB_2$ powders. In this work, $MgB_2$ powders were compacted by two rotating rolls and squeezed out as a form of $MgB_2$ sheets of ~1 mm thickness. The rolling speed of 0.3-0.7 rpm and the gap distance of 0.3-0.8 mm between the two rollers were carefully controlled to get a full compaction of the powders into bulk $MgB_2$ sheets. The densities of $MgB_2$ sheets were 1.98-2.05 g/$cm^3$, which is 75.44-77.99 % of the theoretical value of 2.63 g/$cm^3$. And the density comparison was made compared to those of typical $MgB_2$ bulks from uni-axial pressing and $MgB_2$ wires from Powder-In-Tube processing.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.3
• /
• pp.89-101
• /
• 2014

Recently, as domestic air traffic dramatically increases, the need of ATC(air traffic control) systems has grown for safe and efficient ATM(air traffic management). Especially, for smooth ATC, it is far more important that performance of display system which should show all air traffic situation in FIR(Flight Information Region) without additional latency is guaranteed. In this paper, we design a ASTERIX(All purpose STructured Eurocontrol suRveillance Information eXchange) parsing module to promote stable ATC by minimizing system loads, which is connected with reducing overheads arisen when we parse ASTERIX message. Our ASTERIX parsing module based on pattern matching creates patterns by analyzing received ASTERIX data, and handles following received ASTERIX data using pre-defined procedure through patterns. This module minimizes display errors by rapidly extracting only necessary information for display different from existing parsing module containing unnecessary parsing procedure. Therefore, this designed module is to enable controllers to operate stable ATC. The comparison with existing general bit level ASTERIX parsing module shows that ASTERIX parsing module based on pattern matching has shorter processing delay, higher throughput, and lower CPU usage.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.2
• /
• pp.33-46
• /
• 2011

Since an e-mail has been an important mean of communication and information sharing, there have been much effort to classify e-mails efficiently by their contents. An e-mail has various forms in length and style, and words used in an e-mail are usually irregular. In addition, the criteria of an e-mail classification are subjective. As a result, it is quite difficult for the conventional text classification technique to be adapted to an e-mail classification efficiently. An e-mail classification technique in a commercial e-mail program uses a simple text filtering technique in an e-mail client. In the previous studies on automatic classification of an e-mail, the Naive Bayesian technique based on the probability has been used to improve the classification accuracy, and most of them are on an e-mail in English. This paper proposes the personalized recommendation technique of an email in Korean using a data mining technique of frequent patterns. The proposed technique consists of two phases such as the pre-processing of e-mails in an e-mail folder and the generating a profile for the e-mail folder. The generated profile is used for an e-mail to be classified into the most appropriate e-mail folder by the subjective criteria. The e-mail classification system is also implemented, which adapts the proposed technique.

• Journal of Digital Contents Society
• /
• v.18 no.8
• /
• pp.1593-1601
• /
• 2017

In a big data platform, association rule mining applications could bring some benefits. For instance, in a agricultural big data platform, the association rule mining application could recommend specific products for farmers to grow, which could increase income. The key process of the association rule mining is the frequent itemsets mining, which finds sets of products accompanying together frequently. Former researches about this issue, e.g. Apriori, are not satisfying enough because huge possible sets can cause memory to be overloaded. In order to deal with it, SON algorithm has been proposed, which divides the considered set into many smaller ones and handles them sequently. But in a single machine, SON algorithm cause heavy time consuming. In this paper, we present a method to find association rules in our Hadoop based big data platform, by parallelling SON algorithm. The entire process of association rule mining including pre-processing, SON algorithm based frequent itemset mining, and association rule finding is implemented on Hadoop based big data platform. Through the experiment with real dataset, it is conformed that the proposed method outperforms a brute force method.