• Journal of Information Processing Systems
• /
• 제16권1호
• /
• pp.6-29
• /
• 2020

Biometrics identification using multiple modalities has attracted the attention of many researchers as it produces more robust and trustworthy results than single modality biometrics. In this paper, we present a novel multimodal recognition system that trains a deep learning network to automatically learn features after extracting multiple biometric modalities from a single data source, i.e., facial video clips. Utilizing different modalities, i.e., left ear, left profile face, frontal face, right profile face, and right ear, present in the facial video clips, we train supervised denoising auto-encoders to automatically extract robust and non-redundant features. The automatically learned features are then used to train modality specific sparse classifiers to perform the multimodal recognition. Moreover, the proposed technique has proven robust when some of the above modalities were missing during the testing. The proposed system has three main components that are responsible for detection, which consists of modality specific detectors to automatically detect images of different modalities present in facial video clips; feature selection, which uses supervised denoising sparse auto-encoders network to capture discriminative representations that are robust to the illumination and pose variations; and classification, which consists of a set of modality specific sparse representation classifiers for unimodal recognition, followed by score level fusion of the recognition results of the available modalities. Experiments conducted on the constrained facial video dataset (WVU) and the unconstrained facial video dataset (HONDA/UCSD), resulted in a 99.17% and 97.14% Rank-1 recognition rates, respectively. The multimodal recognition accuracy demonstrates the superiority and robustness of the proposed approach irrespective of the illumination, non-planar movement, and pose variations present in the video clips even in the situation of missing modalities.

• 지능정보연구
• /
• 제21권4호
• /
• pp.93-110
• /
• 2015

기존의 협업필터링 추천시스템 연구는 상품에 대한 고객의 평점(rating)이나 구매 여부 데이터로부터 하나의 프로파일을 생성하고 이를 기반으로 추천 성능을 향상시킬 수 있는 새로운 알고리즘을 개발하는 위주로 진행되어 왔다. 그러나 빅데이터 환경이 도래하면서 기업이 수집할 수 있는 고객 데이터가 풍부해지고 다양해짐에 따라, 보다 정확하게 고객의 선호도나 행태를 파악하는 것이 가능하게 되었고 이러한 데이터, 즉 퍼스널 빅데이터(personal big data)를 추천시스템에 활용하는 연구의 필요성이 대두되고 있다. 본 연구에서는 마케팅의 시장세분화 이론에 근거하여 퍼스널 빅데이터로부터 고객의 선호도나 행태를 다양한 관점에서 표현할 수 있는 5종의 다중 프로파일(multimodal profile)을 개발하고, 이를 활용하여 협업필터링 추천시스템의 성능을 개선하고자 한다. 제안하는 5종의 다중 프로파일은 프로파일 통합 유사도, 개별 프로파일 유사도 평균, 개별 프로파일 유사도 가중 평균이라는 세 가지 앙상블 기법을 통해 협업필터링의 이웃(neighborhood) 탐색과정에 적용된다. 실제 퍼스널 빅데이터에 본 연구에서 제안하는 방법론을 적용한 결과, 단일 프로파일을 사용하는 협업필터링 알고리즘보다 추천 성능이 상당히 개선되었으며 앙상블 방법 중에서는 개별 프로파일 유사도 가중 평균 기법이 가장 높은 추천 성능을 보여주었다. 본 연구는 빅데이터 환경에서 추천시스템을 개발하고자 할 때, 어떠한 성격의 데이터로부터 고객의 특성을 규명하는 프로파일을 만들고 이를 어떻게 결합하여 사용하는 것이 효과적인 지 처음으로 제안하였다는 점에서 그 의의가 있다.

• Asia pacific journal of information systems
• /
• 제33권3호
• /
• pp.737-767
• /
• 2023

With the availability of big customer data and advances in machine learning techniques, the prediction of customer behavior at the session-level has attracted considerable attention from marketing practitioners and scholars. This study aims to predict customer purchase conversion at the session-level by employing customer profile, transaction, and clickstream data. For this purpose, we develop a multimodal deep learning fusion model with dynamic and static features (i.e., DS-fusion). Specifically, we base page views within focal visist and recency, frequency, monetary value, and clumpiness (RFMC) for dynamic and static features, respectively, to comprehensively capture customer characteristics for buying behaviors. Our model with deep learning architectures combines these features for conversion prediction. We validate the proposed model using real-world e-commerce data. The experimental results reveal that our model outperforms unimodal classifiers with each feature and the classical machine learning models with dynamic and static features, including random forest and logistic regression. In this regard, this study sheds light on the promise of the machine learning approach with the complementary method for different modalities in predicting customer behaviors.

• 한국통신학회논문지
• /
• 제27권3A호
• /
• pp.205-210
• /
• 2002

In a Doppler weather radar, high resolution windspeed profile measurements are needed to provide the reliable detection of hazardous weather conditions. For this purpose, the pulse pair method is generally considered to be the most efficient estimator, However, this estimator has some bias errors due to asymmetric spectra and may yield meaningless results in the case of a multimodal return spectrum. Although the poly-pulse pair method can reduce the bias errors of skewed weather spectra, the modes of spectrum may provide more reliable information than the statistical mean for the case of a multimodal or seriously skewed spectrum. Therefore, the idea of relatively simple mode estimator for a weather radar is developed in this paper, Performance simulations show promising results in the detection of hazardous weather conditions.

• Wind and Structures
• /
• 제30권1호
• /
• pp.85-97
• /
• 2020

The accurate estimation of the buffeting response of a bridge pylon is related to the quality of the bridge construction. To evaluate the influence of wind field characteristics on the buffeting response of a pylon in a trumpet-shaped mountain pass, this paper deduced a multimodal coupled buffeting frequency domain calculation method for a variable-section bridge tower under the twisted wind profile condition based on quasi-steady theory. Through the long-term measurement of the wind field of the trumpet-shaped mountain pass, the wind characteristics were studied systematically. The effects of the wind characteristics, wind yaw angles, mean wind speeds, and wind profiles on the buffeting response were discussed. The results show that the mean wind characteristics are affected by the terrain and that the wind profile is severely twisted. The optimal fit distribution of the monthly and annual maximum wind speeds is the log-logistic distribution, and the generalized extreme value I distribution may underestimate the return wind speed. The design wind characteristics will overestimate the buffeting response of the pylon. The buffeting response of the pylon is obviously affected by the wind yaw angle and mean wind speed. To accurately estimate the buffeting response of the pylon in an actual construction, it is necessary to consider the twisted effect of the wind profile.

• 한국통신학회논문지
• /
• 제19권3호
• /
• pp.533-542
• /
• 1994

가상용 도플러 레이다의 경우 기상이변 등을 탐지하기위해서는 대상지역의 강우량, 풍속의 변화, turbulence정도 등을 거리 및 방위각별로 세밀하게 측정, 표시하여줄 수 있어야 한다. 이러한 목적으로 쓰여지고 있는 알고리즘으로 pulse pair추정방법이 가장 효율적인 것으로 인정되어지고 있다. 그러나 이 방법은 레이다 반사신호의 스펙트럼이 비대칭일 경우 bias오차가 크게 생길 수 있으며 두개이상의 peak 점을 갖는 스펙트럼의 경우에는 무의미한 결과가 얻어질 수 있다. 따라서 본 논문에서는 이와 같은 bias 오차에 대해서 분석하였으며 이러한 오차를 줄이기 위한 개선된 방법을 제시하였다. 또한 여러개의 peak점을 갖는 스펙트럼이나 비대칭의 정도가 심할 경우 mode 추정 방법을 이용한 탐지 기법에 대해서도 연구, 검토되어졌다.

• Hip & pelvis
• /
• 제36권2호
• /
• pp.108-119
• /
• 2024

Purpose: The incidence of deep vein thrombosis (DVT) following total hip arthroplasty (THA) without chemoprophylaxis could be as high as 50% in Caucasians. However, according to several subsequent studies, the incidence of venous thromboembolic events (VTE) in Asians was much lower. The routine use of chemoprophylaxis, which could potentially cause increased bleeding, infection, and wound complications, has been questioned in low-incidence populations. The objective of this study is to determine the incidence of VTE after primary THA without chemoprophylaxis in an Asian population using a fast-track rehabilitation protocol and to verify the safety profile for use of 'mechanical prophylaxis alone' in patients with standard risk of VTE. Materials and Methods: This is a retrospective cohort study of 542 Hong Kong Chinese patients who underwent primary THA without chemoprophylaxis. All patients received intermittent pneumatic compression and graduated compression stockings as mechanical prophylaxis. Multimodal pain management was applied in order to facilitate early mobilisation. Routine duplex ultrasonography was performed between the fourth and seventh postoperative day for detection of proximal DVT. Results: All patients were Chinese (mean age, 63.0±11.9 years). Six patients developed proximal DVT (incidence rate, 1.1%). None of the patients had symptomatic or fatal pulmonary embolism. Conclusion: The incidence of VTE after primary THA without chemical prophylaxis can be low in Asian populations when following a fast-track rehabilitation protocol. Mechanical prophylaxis alone can be regarded as a reasonably safe practice in terms of a balanced benefit-to-risk ratio for Asian patients with standard risk of VTE.

• Journal of Digestive Cancer Research
• /
• 제6권1호
• /
• pp.25-31
• /
• 2018

Background: In Korea, stomach cancer is the second most common malignancy and the third leading cause of cancer-related deaths. the time of diagnosis is very important for treatment so early detection and surgery are currently considered the mainstay of treatment, when diagnosed advanced with tumor extension through the gastric wall and direct extension into other organs, with metastatic involvement. Recently, new drugs, drug combinations, and multimodal approaches have been used to treat this disease and In cancers over expressing or amplifying HER2, the combination of cisplatin-fluoropyrimidine-trastuzumab is considered to be the treatment of reference. but At present, the choice of treatment schedule for HER2-negative tumors is based on the medical institution's preferences and adverse effects profile. The aim of this study was to evaluate the effectiveness and safety of using FOLFOX regimen as a first-line therapy or a salvage therapy in the patients with HER2-negative advanced or metastatic gastric cancer. Methods: We retrospective reviewed the patient medical record from March 2012 to July 2017. This study evaluated 113 patients. Sixty-eight patients were treated with the FOLFOX regimen for the first time (first-line group) and 45 patients were treated with the FOLFOX regimen as a second (35 patients) or third (10 patients) chemotherapy (salvage group). Results: In the first-line group, the response rate was 54.9%. In the salvage therapy group, the response rate was 24.4% and The difference was statistically significant (p=0.205). The median TTP of the first-line group was 10.7 months (95% confidence interval [95% CI], 7.8-13.7 months) and that of salvage line group was 6.1 months (95% CI, 3.8-8.4 months). The median OS of the first-line group was 15.8 months (95% CI, 12.7-18.9 months) and that of the salvage therapy group was 10.2 months (95% CI, 8.2-11.9 months). drug toxicity was similar andtolerable between two groups. Conclusion: In patients with unresctable metastatic gastric cancer, after failing to respond to first-line therapy, most patients have no alternative other than second-line therapy because the disease is highly progressive. if the performance status of the patient is good enough to be eligible to treatments beyond best supportive care. FOLFOX regimen can be a considerable therapeutic option for salvage treatment.

• 한국해양학회지:바다
• /
• 제4권2호
• /
• pp.117-126
• /
• 1999

해안침식이 우세한 반폐쇄적 조간대의 퇴적작용을 규명하기 위해 우리 나라 서해안의 함평만 조간대에서 지형, 퇴적물, 집적율, 해안절벽의 침식율 등을 조사하였다. 함평만 조간대는 전반적으로 조류세곡(tidal creek)의 발달이 미약하고, 고조선 해빈(high-tide beach), 조간대 사주(intertidal sand shoal) 등에 의해 지형적 기복이 표시되며, 대체로 상부방향으로 오목한 지형단면을 보인다. 조간대 퇴적물은 해안절벽의 침식에 의해 공급된 육상기원의 조립 퇴적물(자갈, 모래)과 조석작용에 의해 바다로부터 공급된 세립 퇴적물(펄)이 다양하게 혼합됨으로써 복모드형 입도분포, 불량한 분급, 양의 왜도 등의 조직적 특성을 보인다. 표층 퇴적물의 퇴적상 분포는 대체로 저조선 방향과 만 입구 방향으로 세립해지는 경향을 보여, 조간대의 일반적 경향과 다르다. 조간대층은 대부분 생물교란에 의해 균질화된 상태이며, 만 입구의 펄질 조간대에서 제한적으로 렌즈상층리(lenticular bedding), 평행엽리(parallel lamination) 통이 관찰된다. 조간대의 침식은 상부 조간대와 하부 조간대를 중심으로 추로 봄철(3~5 월)과 겨울철(10~3월)에 우세하며, 침식우세 지역의 경우 연평균 침식율이 5.2 cm/yr에 이른다. 침식에 의한 해안절벽의 후퇴는 내만역의 절벽에서 최고 1.4 m/yr의 속도로 이루어지는데, 대체로 초여름(5~6 월)에 우세하여 조간대의 최대 침식시기(봄철)와 1~2 개월의 시간차를 보인다. 이렇듯 함평만과 같은 전형적인 반폐쇄적 환경 내에서 조간대와 해안절벽이 침식되는 까닭은 풍파와 지역적 해수면 상승의 상호작용 때문인 것으로 판단된다.