• 전자공학회논문지
• /
• 제51권2호
• /
• pp.23-29
• /
• 2014

지능형 교통 시스템(Intelligent Transportation Systems)의 첨단 교통 관리 시스템(Advanced Traffic Management System)은 고화질 카메라, 고성능 레이더 센서와 같은 향상된 인프라를 통하여 도로 상의 차량 속도, 통행량, 돌발 상황 등의 교통 상황을 실시간으로 분석하며 관련 업무를 자동화하고 있다. 특히 도로 이용자의 안전을 위해서는 돌발 상황 자동 검지 및 2차 사고 방지를 위한 시스템이 필요하다. 이러한 유고 검지 및 관리 시스템에서는 CCTV 기반 영상 검지와 레이더를 이용한 물체검지가 주로 사용된다. 본 논문은 다중 감시용 카메라를 사용한 실시간 고속도로 돌발 상황 검지 시스템에서 모자이크(mosaic) 동영상을 구성하는 방법과 다양한 각도에서 촬영된 움직이는 객체를 보다 정확하게 추적할 수 있는 배경 모델링에 기반한 알고리즘을 제안하였다. 실험결과 영상검지는 레이더검지의 근거리 음영 영역과 원거리 검지한계 영역을 보완해 줄 수 있을 뿐만 아니라 악천후를 제외한 주간 검지에서 보다 나은 분류 특징들을 갖고 있음을 확인 할 수 있었다.

• 대한전자공학회논문지SP
• /
• 제45권4호
• /
• pp.107-114
• /
• 2008

본 논문은 남/여 성별에 기반해 음성을 평상, 기쁨, 슬픔, 화남의 4가지 감성 상태로 분류하는 감성인식 시스템을 구축하였다. 제안된 시스템은 입력 음성으로부터 1차적으로 남/여 성별을 분류하고, 분류된 성별을 기반으로 남/여 각기 최적의 특징벡터 열을 적용하여 감성인식을 수행함으로써 감성인식 성공률을 향상시켰다. 또한 음성인식에서 주로 사용되는 ZCPA(Zero Crossings with Peak Amplitudes)를 감성인식용 특징벡터로 사용하여 성능을 향상시켰으며, 남/여 각각의 특징 벡터 열을 최적화하기 위해 SFS(Sequential Forward Selection) 기법을 사용하였다. 감성 패턴 분류기로는 k-NN과 SVM을 비교하여 실험하였다. 실험결과 제안 시스템은 4가지 감성상태에 대해 약 85.3%의 높은 감성 인식 성공률을 달성할 수 있어 향후 감성을 인식하는 콜센터, 휴머노이드형 로봇이나 유비쿼터스(Ubiquitous) 환경 등 다양한 분야에서 감성인식 정보를 유용하게 사용될 수 있을 것으로 기대된다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권8호
• /
• pp.2881-2894
• /
• 2014

In complicated environment, context information plays an important role in image segmentation/labeling. The recently proposed auto-context algorithm is one of the effective context-based methods. However, the standard auto-context approach samples the context locations utilizing a fixed radius sequence, which is sensitive to large scale-change of objects. In this paper, we present a scale invariant auto-context (SIAC) algorithm which is an improved version of the auto-context algorithm. In order to achieve scale-invariance, we try to approximate the optimal scale for the image in an iterative way and adopt the corresponding optimal radius sequence for context location sampling, both in training and testing. In each iteration of the proposed SIAC algorithm, we use the current classification map to estimate the image scale, and the corresponding radius sequence is then used for choosing context locations. The algorithm iteratively updates the classification maps, as well as the image scales, until convergence. We demonstrate the SIAC algorithm on several image segmentation/labeling tasks. The results demonstrate improvement over the standard auto-context algorithm when large scale-change of objects exists.

• International Journal of Internet, Broadcasting and Communication
• /
• 제13권3호
• /
• pp.92-103
• /
• 2021

With the continuous acceleration of economic and social development, people gradually pay attention to their health, improve their living environment, diet, strengthen exercise, and even conduct regular health examination, to ensure that they always understand the health status. Even so, people still face many health problems, and the number of chronic diseases is increasing. Recently, COVID-19 has also reminded people that public health problems are also facing severe challenges. With the development of artificial intelligence equipment and technology, medical diagnosis expert systems based on big data have become a topic of concern to many researchers. At present, there are many algorithms that can help computers initially diagnose diseases for patients, but they want to improve the accuracy of diagnosis. And taking into account the pathology that varies from person to person, the health diagnosis expert system urgently needs a new algorithm to improve accuracy. Through the understanding of classic algorithms, this paper has optimized it, and finally proved through experiments that the combined classification algorithm improved by latent factors can meet the needs of medical intelligent diagnosis.

• International Journal of Control, Automation, and Systems
• /
• 제4권6호
• /
• pp.705-713
• /
• 2006

For the analysis of multiunit extracellular neural signals as multiple spike trains, neural spike sorting is essential. Existing algorithms for the spike sorting have been unsatisfactory when the signal-to-noise ratio(SNR) is low, especially for implementation of fully-automated systems. We present a novel method that shows satisfactory performance even under low SNR, and compare its performance with a recent method based on principal component analysis(PCA) and fuzzy c-means(FCM) clustering algorithm. Our system consists of a spike detector that shows high performance under low SNR, a feature extractor that utilizes projection pursuit based on negentropy maximization, and an unsupervised classifier based on Gaussian mixture model. It is shown that the proposed feature extractor gives better performance compared to the PCA, and the proposed combination of spike detector, feature extraction, and unsupervised classification yields much better performance than the PCA-FCM, in that the realization of fully-automated unsupervised spike sorting becomes more feasible.

• 한국멀티미디어학회논문지
• /
• 제13권6호
• /
• pp.841-848
• /
• 2010

Conventional filters using email header and body information equally judge whether an incoming email is spam or not. However this is unrealistic in everyday life because each person has different criteria to judge what is spam or not. To resolve this problem, we consider user preference information as well as email category information derived from the email content. In this paper, we have developed a personalized anti-spam system using ontologies constructed from rules derived in a data mining process. The reason why traditional content-based filters are not applicable to the proposed experimental situation is described. In also, several experiments constructing classifiers to decide email category and comparing classification rule learners are performed. Especially, an ID3 decision tree algorithm improved the overall accuracy around 17% compared to a conventional SVM text miner on the decision of email category. Some discussions about the axioms generated from the experimental dataset are given too.

• 대한음성학회지:말소리
• /
• 제57호
• /
• pp.139-152
• /
• 2006

In this paper, we present a phrase break prediction method using CRF(Conditional Random Fields), which has good performance at classification problems. The phrase break prediction problem was mapped into a classification problem in our research. We trained the CRF using the various linguistic features which was extracted from POS(Part Of Speech) tag, lexicon, length of word, and location of word in the sentences. Combined linguistic features were used in the experiments, and we could collect some linguistic features which generate good performance in the phrase break prediction. From the results of experiments, we can see that the proposed method shows improved performance on previous methods. Additionally, because the linguistic features are independent of each other in our research, the proposed method has higher flexibility than other methods.

• The Journal of the Acoustical Society of Korea
• /
• 제25권2E호
• /
• pp.74-78
• /
• 2006

This paper proposes an effective method for classifying emotions of the music from its acoustical signals. Two feature sets, timbre and tempo, are directly extracted from the modified discrete cosine transform coefficients (MDCT), which are the output of partial MP3 (MPEG 1 Layer 3) decoder. Our tempo feature extraction method is based on the long-term modulation spectrum analysis. In order to effectively combine these two feature sets with different time resolution in an integrated system, a classifier with two layers based on AdaBoost algorithm is used. In the first layer the MDCT-driven timbre features are employed. By adding the MDCT-driven tempo feature in the second layer, the classification precision is improved dramatically.

• 음성과학
• /
• 제8권2호
• /
• pp.61-71
• /
• 2001

Speech material, which is collected from ARS(Automatic Response System), was analyzed and classified into disease and non-disease state. The material include 11 different kinds of diseases. Along with ARS speech, DAT(Digital Audio Tape) speech is collected in parallel to give the bench mark. To analyze speech material, analysis tools, which is developed local laboratory, are used to provide an improved and robust performance to the obtained parameters. To classify speech into disease and non-disease class, multi-layered neural network was used. Three different combinations of 3, 6, 12 parameters are tested to obtain the proper network size and to find the best performance. From the experiment, the classification rate of 92.5% was obtained.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2022년도 추계학술발표대회
• /
• pp.493-494
• /
• 2022

The development of livestock faces intensive farming results in a rising need for recognition of individual animals such as cows and pigs is related to high traceability. In this paper, we present a non-invasive biometrics systematic approach based on the deep-learning classification model to pig face identification. Firstly, in our systematic method, we build a ROS data collection system block to collect 10 pig face data images. Secondly, we proposed a preprocessing block in that we utilize the SSIM method to filter some images of collected images that have high similarity. Thirdly, we employ the improved image classification model of CNN (ViT), which uses the finetuning and pretraining technique to recognize the individual pig face. Finally, our proposed method achieves the accuracy about 98.66%.