• Journal of Intelligence and Information Systems
• /
• v.24 no.2
• /
• pp.21-35
• /
• 2018

Since the introduction of MP3 players, CD recordings have gradually been vanishing, and the music consuming environment of music users is shifting to mobile devices. The introduction of smart devices has increased the utilization of music through music playback, mass storage, and search functions that are integrated into smartphones and tablets. At the time of initial MP3 player supply, the bitrate of the compressed music contents generally was 128 Kbps. However, as increasing of the demand for high quality music, sound quality of 384 Kbps appeared. Recently, music content of FLAC (Free License Audio Codec) format using lossless compression method is becoming popular. The download service of many music sites in Korea has classified by unlimited download with technical protection and limited download without technical protection. Digital Rights Management (DRM) technology is used as a technical protection measure for unlimited download, but it can only be used with authenticated devices that have DRM installed. Even if music purchased by the user, it cannot be used by other devices. On the contrary, in the case of music that is limited in quantity but not technically protected, there is no way to enforce anyone who distributes it, and in the case of high quality music such as FLAC, the loss is greater. In this paper, the author proposes an audio watermarking technology for copyright protection of high quality stereo music. Two kinds of information, "Copyright" and "Copy_free", are generated by using the turbo code. The two watermarks are composed of 9 bytes (72 bits). If turbo code is applied for error correction, the amount of information to be inserted as 222 bits increases. The 222-bit watermark was expanded to 1024 bits to be robust against additional errors and finally used as a watermark to insert into stereo music. Turbo code is a way to recover raw data if the damaged amount is less than 15% even if part of the code is damaged due to attack of watermarked content. It can be extended to 1024 bits or it can find 222 bits from some damaged contents by increasing the probability, the watermark itself has made it more resistant to attack. The proposed algorithm uses quantization in DCT so that watermark can be detected efficiently and SNR can be improved when stereo music is converted into mono. As a result, on average SNR exceeded 40dB, resulting in sound quality improvements of over 10dB over traditional quantization methods. This is a very significant result because it means relatively 10 times improvement in sound quality. In addition, the sample length required for extracting the watermark can be extracted sufficiently if the length is shorter than 1 second, and the watermark can be completely extracted from music samples of less than one second in all of the MP3 compression having a bit rate of 128 Kbps. The conventional quantization method can extract the watermark with a length of only 1/10 compared to the case where the sampling of the 10-second length largely fails to extract the watermark. In this study, since the length of the watermark embedded into music is 72 bits, it provides sufficient capacity to embed necessary information for music. It is enough bits to identify the music distributed all over the world. 272 can identify $4*10^{21}$, so it can be used as an identifier and it can be used for copyright protection of high quality music service. The proposed algorithm can be used not only for high quality audio but also for development of watermarking algorithm in multimedia such as UHD (Ultra High Definition) TV and high-resolution image. In addition, with the development of digital devices, users are demanding high quality music in the music industry, and artificial intelligence assistant is coming along with high quality music and streaming service. The results of this study can be used to protect the rights of copyright holders in these industries.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• Journal of Intelligence and Information Systems
• /
• v.22 no.4
• /
• pp.177-192
• /
• 2016

Machine learning is a field of artificial intelligence. It refers to an area of computer science related to providing machines the ability to perform their own data analysis, decision making and forecasting. For example, one of the representative machine learning models is artificial neural network, which is a statistical learning algorithm inspired by the neural network structure of biology. In addition, there are other machine learning models such as decision tree model, naive bayes model and SVM(support vector machine) model. Among the machine learning models, we use SVM model in this study because it is mainly used for classification and regression analysis that fits well to our study. The core principle of SVM is to find a reasonable hyperplane that distinguishes different group in the data space. Given information about the data in any two groups, the SVM model judges to which group the new data belongs based on the hyperplane obtained from the given data set. Thus, the more the amount of meaningful data, the better the machine learning ability. In recent years, many financial experts have focused on machine learning, seeing the possibility of combining with machine learning and the financial field where vast amounts of financial data exist. Machine learning techniques have been proved to be powerful in describing the non-stationary and chaotic stock price dynamics. A lot of researches have been successfully conducted on forecasting of stock prices using machine learning algorithms. Recently, financial companies have begun to provide Robo-Advisor service, a compound word of Robot and Advisor, which can perform various financial tasks through advanced algorithms using rapidly changing huge amount of data. Robo-Adviser's main task is to advise the investors about the investor's personal investment propensity and to provide the service to manage the portfolio automatically. In this study, we propose a method of forecasting the Korean volatility index, VKOSPI, using the SVM model, which is one of the machine learning methods, and applying it to real option trading to increase the trading performance. VKOSPI is a measure of the future volatility of the KOSPI 200 index based on KOSPI 200 index option prices. VKOSPI is similar to the VIX index, which is based on S&P 500 option price in the United States. The Korea Exchange(KRX) calculates and announce the real-time VKOSPI index. VKOSPI is the same as the usual volatility and affects the option prices. The direction of VKOSPI and option prices show positive relation regardless of the option type (call and put options with various striking prices). If the volatility increases, all of the call and put option premium increases because the probability of the option's exercise possibility increases. The investor can know the rising value of the option price with respect to the volatility rising value in real time through Vega, a Black-Scholes's measurement index of an option's sensitivity to changes in the volatility. Therefore, accurate forecasting of VKOSPI movements is one of the important factors that can generate profit in option trading. In this study, we verified through real option data that the accurate forecast of VKOSPI is able to make a big profit in real option trading. To the best of our knowledge, there have been no studies on the idea of predicting the direction of VKOSPI based on machine learning and introducing the idea of applying it to actual option trading. In this study predicted daily VKOSPI changes through SVM model and then made intraday option strangle position, which gives profit as option prices reduce, only when VKOSPI is expected to decline during daytime. We analyzed the results and tested whether it is applicable to real option trading based on SVM's prediction. The results showed the prediction accuracy of VKOSPI was 57.83% on average, and the number of position entry times was 43.2 times, which is less than half of the benchmark (100 times). A small number of trading is an indicator of trading efficiency. In addition, the experiment proved that the trading performance was significantly higher than the benchmark.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.71-88
• /
• 2017

Language models were originally developed for speech recognition and language processing. Using a set of example sentences, a language model predicts the next word or character based on sequential input data. N-gram models have been widely used but this model cannot model the correlation between the input units efficiently since it is a probabilistic model which are based on the frequency of each unit in the training set. Recently, as the deep learning algorithm has been developed, a recurrent neural network (RNN) model and a long short-term memory (LSTM) model have been widely used for the neural language model (Ahn, 2016; Kim et al., 2016; Lee et al., 2016). These models can reflect dependency between the objects that are entered sequentially into the model (Gers and Schmidhuber, 2001; Mikolov et al., 2010; Sundermeyer et al., 2012). In order to learning the neural language model, texts need to be decomposed into words or morphemes. Since, however, a training set of sentences includes a huge number of words or morphemes in general, the size of dictionary is very large and so it increases model complexity. In addition, word-level or morpheme-level models are able to generate vocabularies only which are contained in the training set. Furthermore, with highly morphological languages such as Turkish, Hungarian, Russian, Finnish or Korean, morpheme analyzers have more chance to cause errors in decomposition process (Lankinen et al., 2016). Therefore, this paper proposes a phoneme-level language model for Korean language based on LSTM models. A phoneme such as a vowel or a consonant is the smallest unit that comprises Korean texts. We construct the language model using three or four LSTM layers. Each model was trained using Stochastic Gradient Algorithm and more advanced optimization algorithms such as Adagrad, RMSprop, Adadelta, Adam, Adamax, and Nadam. Simulation study was done with Old Testament texts using a deep learning package Keras based the Theano. After pre-processing the texts, the dataset included 74 of unique characters including vowels, consonants, and punctuation marks. Then we constructed an input vector with 20 consecutive characters and an output with a following 21st character. Finally, total 1,023,411 sets of input-output vectors were included in the dataset and we divided them into training, validation, testsets with proportion 70:15:15. All the simulation were conducted on a system equipped with an Intel Xeon CPU (16 cores) and a NVIDIA GeForce GTX 1080 GPU. We compared the loss function evaluated for the validation set, the perplexity evaluated for the test set, and the time to be taken for training each model. As a result, all the optimization algorithms but the stochastic gradient algorithm showed similar validation loss and perplexity, which are clearly superior to those of the stochastic gradient algorithm. The stochastic gradient algorithm took the longest time to be trained for both 3- and 4-LSTM models. On average, the 4-LSTM layer model took 69% longer training time than the 3-LSTM layer model. However, the validation loss and perplexity were not improved significantly or became even worse for specific conditions. On the other hand, when comparing the automatically generated sentences, the 4-LSTM layer model tended to generate the sentences which are closer to the natural language than the 3-LSTM model. Although there were slight differences in the completeness of the generated sentences between the models, the sentence generation performance was quite satisfactory in any simulation conditions: they generated only legitimate Korean letters and the use of postposition and the conjugation of verbs were almost perfect in the sense of grammar. The results of this study are expected to be widely used for the processing of Korean language in the field of language processing and speech recognition, which are the basis of artificial intelligence systems.

• Journal of Intelligence and Information Systems
• /
• v.17 no.3
• /
• pp.25-41
• /
• 2011

Recently, support vector machines (SVMs) are being recognized as competitive tools as compared with other data mining techniques for solving pattern recognition or classification decision problems. Furthermore, many researches, in particular, have proved them more powerful than traditional artificial neural networks (ANNs) (Amendolia et al., 2003; Huang et al., 2004, Huang et al., 2005; Tay and Cao, 2001; Min and Lee, 2005; Shin et al., 2005; Kim, 2003).The classification decision, such as a binary or multi-class decision problem, used by any classifier, i.e. data mining techniques is so cost-sensitive particularly in financial classification problems such as the credit ratings that if the credit ratings are misclassified, a terrible economic loss for investors or financial decision makers may happen. Therefore, it is necessary to convert the outputs of the classifier into wellcalibrated posterior probabilities-based multiclass credit ratings according to the bankruptcy probabilities. However, SVMs basically do not provide such probabilities. So it required to use any method to create the probabilities (Platt, 1999; Drish, 2001). This paper applied AdaBoost algorithm-based support vector machines (SVMs) into a bankruptcy prediction as a binary classification problem for the IT companies in Korea and then performed the multi-class credit ratings of the companies by making a normal distribution shape of posterior bankruptcy probabilities from the loss functions extracted from the SVMs. Our proposed approach also showed that their methods can minimize the misclassification problems by adjusting the credit grade interval ranges on condition that each credit grade for credit loan borrowers has its own credit risk, i.e. bankruptcy probability.

• Journal of Korean Society for Geospatial Information Science
• /
• v.8 no.1 s.15
• /
• pp.51-63
• /
• 2000

Dynamic traffic assignment(DTA) has been a topic of substantial research during the past decade. While DTA is gradually maturing, many aspects of DTA still need improvement, especially regarding its formulation and solution algerian Recently, with its promise for In(Intelligent Transportation System) and GIS(Geographic Information System) applications, DTA have received increasing attention. This potential also implies higher requirement for DTA modeling, especially regarding its solution efficiency for real-time implementation. But DTA have many mathematical difficulties in searching process due to the complexity of spatial and temporal variables. Although many solution algorithms have been studied, conventional methods cannot iud the solution in case that objective function or constraints is not convex. In this paper, the genetic algorithm to find the solution of DTA is applied and the Merchant-Nemhauser model is used as DTA model because it has a nonconvex constraint set. To handle the nonconvex constraint set the GENOCOP III system which is a kind of the genetic algorithm is used in this study. Results for the sample network have been compared with the results of conventional method.

• Journal of The Korean Association For Science Education
• /
• v.38 no.2
• /
• pp.219-234
• /
• 2018

In this study, we explored the possibility of automating the process of analyzing elements of scientific argument in the context of a Korean classroom. To gather training data, we collected 990 sentences from science education journals that illustrate the results of coding elements of argumentation according to Toulmin's argumentation structure framework. We extracted 483 sentences as a test data set from the transcription of students' discourse in scientific argumentation activities. The words and morphemes of each argument were analyzed using the Python 'KoNLPy' package and the 'Kkma' module for Korean Natural Language Processing. After constructing the 'argument-morpheme:class' matrix for 1,473 sentences, five machine learning techniques were applied to generate predictive models relating each sentences to the element of argument with which it corresponded. The accuracy of the predictive models was investigated by comparing them with the results of pre-coding by researchers and confirming the degree of agreement. The predictive model generated by the k-nearest neighbor algorithm (KNN) demonstrated the highest degree of agreement [54.04% (${\kappa}=0.22$)] when machine learning was performed with the consideration of morpheme of each sentence. The predictive model generated by the KNN exhibited higher agreement [55.07% (${\kappa}=0.24$)] when the coding results of the previous sentence were added to the prediction process. In addition, the results indicated importance of considering context of discourse by reflecting the codes of previous sentences to the analysis. The results have significance in that, it showed the possibility of automating the analysis of students' argumentation activities in Korean language by applying machine learning.

• The Journal of Bigdata
• /
• v.4 no.2
• /
• pp.23-34
• /
• 2019

Recently, the crime that utilizes the digital platform is continuously increasing. About 140,000 cases occurred in 2015 and about 150,000 cases occurred in 2016. Therefore, it is considered that there is a limit handling those online crimes by old-fashioned investigation techniques. Investigators' manual online search and cognitive investigation methods those are broadly used today are not enough to proactively cope with rapid changing civil crimes. In addition, the characteristics of the content that is posted to unspecified users of social media makes investigations more difficult. This study suggests the site-based collection and the Open API among the content web collection methods considering the characteristics of the online media where the infringement crimes occur. Since illegal content is published and deleted quickly, and new words and alterations are generated quickly and variously, it is difficult to recognize them quickly by dictionary-based morphological analysis registered manually. In order to solve this problem, we propose a tokenizing method in the existing dictionary-based morphological analysis through WPM (Word Piece Model), which is a data preprocessing method for quick recognizing and responding to illegal contents posting online infringement crimes. In the analysis of data, the optimal precision is verified through the Vote-based ensemble method by utilizing a classification learning model based on supervised learning for the investigation of illegal contents. This study utilizes a sorting algorithm model centering on illegal multilevel business cases to proactively recognize crimes invading the public economy, and presents an empirical study to effectively deal with social data collection and content investigation.

• Journal of Korea Game Society
• /
• v.9 no.3
• /
• pp.139-150
• /
• 2009

This paper attempts to investigate the establishment of an interactive knowledge base for action planning by virtual agents and an interactive knowledge-based planning system. A fixed knowledge base is unable to properly handle a change in circumstances because fixed planning is only available under a fixed knowledge base. Therefore, this paper proposes the establishment of an interactive knowledge base which is applicable to diverse environments and an artificial intelligence planning system in which an interactive knowledge base is available. The interactive knowledge base proposed in this paper consists of motivation, behavior, object and action. The association relationship between knowledge base and its input is set using an automation tool. With this tool, a user can easily add to or amend the components of the knowledge base. With this knowledge base, a character plans all action items and chooses one of them to take an action. Since a new action can be applicable by updating the knowledge base even when the character environment changes, it is very useful for virtual reality content developers. This paper has established a relationship between scalable interactive knowledge base components and other components and proposes a convenient input tool and a planning system algorithm effective for an interactive knowledge base. The results of this study have been verified through testing in a virtual environment ('virtual library').