• Smart Media Journal
• /
• v.12 no.11
• /
• pp.48-56
• /
• 2023

The technology of Three-dimensional human posture estimation is used in sports, motion recognition, and special effects of video media. Among various methods for this, multi-view 3D human pose estimation is essential for precise estimation even in complex real-world environments. But Existing models for multi-view 3D human posture estimation have the disadvantage of high order of time complexity as they use 3D feature maps. This paper proposes a method to extend an existing monocular viewpoint multi-frame model based on Transformer with lower time complexity to 3D human posture estimation for multi-viewpoints. To expand to multi-viewpoints our proposed method first generates an 8-dimensional joint coordinate that connects 2-dimensional joint coordinates for 17 joints at 4-vieiwpoints acquired using the 2-dimensional human posture detector, CPN(Cascaded Pyramid Network). This paper then converts them into 17×32 data with patch embedding, and enters the data into a transformer model, finally. Consequently, the MLP(Multi-Layer Perceptron) block that outputs the 3D-human posture simultaneously updates the 3D human posture estimation for 4-viewpoints at every iteration. Compared to Zheng[5]'s method the number of model parameters of the proposed method was 48.9%, MPJPE(Mean Per Joint Position Error) was reduced by 20.6 mm (43.8%) and the average learning time per epoch was more than 20 times faster.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.4E
• /
• pp.112-119
• /
• 2004

This paper presents a method for the automatic recognition of pitch accents with no prior knowledge about the phonetic content of the signal (no knowledge of word or phoneme boundaries or of phoneme labels). The recognition algorithm used in this paper is a time-delay recurrent neural network (TDRNN). A TDRNN is a neural network classier with two different representations of dynamic context: delayed input nodes allow the representation of an explicit trajectory F0(t), while recurrent nodes provide long-term context information that can be used to normalize the input F0 trajectory. Performance of the TDRNN is compared to the performance of a MLP (multi-layer perceptron) and an HMM (Hidden Markov Model) on the same task. The TDRNN shows the correct recognition of $91.9{\%}\;of\;pitch\;events\;and\;91.0{\%}$ of pitch non-events, for an average accuracy of $91.5{\%}$ over both pitch events and non-events. The MLP with contextual input exhibits $85.8{\%},\;85.5{\%},\;and\;85.6{\%}$ recognition accuracy respectively, while the HMM shows the correct recognition of $36.8{\%}\;of\;pitch\;events\;and\;87.3{\%}$ of pitch non-events, for an average accuracy of $62.2{\%}$ over both pitch events and non-events. These results suggest that the TDRNN architecture is useful for the automatic recognition of pitch accents.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.7
• /
• pp.50-56
• /
• 1998

본 논문은 자동 음소 분할기의 음소 경계 오류를 보상하기 위한 후처리 (Postprocessing)에 관한 연구이다. 이는 현재 음성 합성을 위한 음성/언어학적 연구, 운율 모델링, 합성단위 자동 생성 연구 등에 대량의 음소 단위 분절과 음소 레이블링된 데이터의 필요성에 따른 연구의 일환이다. 특히 수작업에 의한 분절 및 레이블링은 일관성의 유지가 어렵고 긴 시간이 소요되므로 자동 분절 기술이 더욱 중요시 되고 있다. 따라서, 본 논문은 자동 분절 경계의 오류 범위를 줄일 수 있는 후처리기를 제안하여 자동 분절 결과를 직접 합성 단위로 사용할 수 있고 대량의 합성용 운율 데이터 베이스 구축에 유용함을 기술한다. 제안된 후처리기는 수작업으로 조정된 데이터의 특징 벡터를 다층 신경회로망 (MLP:Multi-layer perceptron)을 통해 학습을 한 후, ETRI(Electronics and Telecommunication Research Institute)에서 개발된 음성 언어 번역 시스템을 이용한 자동 분절 결과와 후처리기인 MLP를 이용하여 새로운 음소 경계를 추출한다. 고립단어로 발성된 합성 데이터베이스에서 후처리기로 보정된 분절 결과는 음성 언어 번역 시스템의 분할율보 다 약 25%의 향상된 성능을 보였으며, 절대 오류(｜Hand label position-Auto label position ｜)는 약 39%가 향상되었다. 이는 MLP를 이용한 후처리기로 자동 분절 오류의 범위를 줄 일 수 있고, 대량의 합성용 운율 데이터 베이스 구축 및 합성 단위의 자동생성에 이용될 수 있음을 보이는 것이다.

• Earthquakes and Structures
• /
• v.15 no.4
• /
• pp.431-441
• /
• 2018

This study provides fragility-based assessment of seismic performance of reinforced concrete bridges. Seismic fragility curves were created using nonlinear analysis (NA) and artificial neural networks (ANNs). Nonlinear response history analyses were performed, in order to calculate the seismic performances of the bridges. To this end, 306 bridge-earthquake cases were considered. A multi-layered perceptron (MLP) neural network was implemented to predict the seismic performances of the selected bridges. The MLP neural networks considered herein consist of an input layer with four input vectors; two hidden layers and an output vector. In order to train ANNs, 70% of the numerical results were selected, and the remained 30% were employed for testing the reliability and validation of ANNs. Several structures of MLP neural networks were examined in order to obtain suitable neural networks. After achieving the most proper structure of neural network, it was used for generating new data. A total number of 600 new bridge-earthquake cases were generated based on neural simulation. Finally, probabilistic seismic safety analyses were conducted. Herein, fragility curves were developed using numerical results, neural predictions and the combination of numerical and neural data. Results of this study revealed that ANNs are suitable tools for predicting seismic performances of RC bridges. It was also shown that yield stresses of the reinforcements is one of the important sources of uncertainty in fragility analysis of RC bridges.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.30 no.1
• /
• pp.39-49
• /
• 2020

In this paper, we show that the Chinese standard block cipher SM4 is vulnerable to the side channel attacks and present a countermeasure to resist them. We firstly validate that the secret key of SM4 can be recovered by differential power analysis(DPA) and correlation power analysis(CPA) attacks. Therefore we analyze the vulnerable element caused by power attack and propose a first order masking-based countermeasure to defeat DPA and CPA attacks. Although the proposed countermeasure unfortunately is still vulnerable to the profiling power attacks such as deep learning-based multi layer perceptron(MLP), it can sufficiently overcome the non-profiling attacks such as DPA and CPA.

• International Journal of Computer Science & Network Security
• /
• v.21 no.3
• /
• pp.103-111
• /
• 2021

Obesity affects individuals of all gender and ages worldwide; consequently, several studies have performed great works to define factors causing it. This study develops an effective method to trace obesity levels based on supervised data mining techniques such as Random Forest and Multi-Layer Perception (MLP), so as to tackle this universal epidemic. Notably, the dataset was from countries like Mexico, Peru, and Colombia in the 14- 61year age group, with varying eating habits and physical conditions. The data includes 2111 instances and 17 attributes labelled using NObesity, which facilitates categorization of data using Overweight Levels l I and II, Insufficient Weight, Normal Weight, as well as Obesity Type I to III. This study found that the highest accuracy was achieved by Random Forest algorithm in comparison to the MLP algorithm, with an overall classification rate of 96.7%.

• Journal of Information Technology Applications and Management
• /
• v.29 no.3
• /
• pp.43-55
• /
• 2022

In this study, with the goal of developing a deep learning-based product recommendation model for effective matching of influencers and products, a deep learning model with a collaborative filtering model combined with generalized matrix decomposition(GMF), a collaborative filtering model based on multi-layer perceptron (MLP), and neural collaborative filtering and generalized matrix Factorization (NeuMF), a hybrid model combining GMP and MLP was developed and tested. In particular, we utilize one-class problem free boosting (OCF-B) method to solve the one-class problem that occurs when training is performed only on positive cases using implicit feedback in the deep learning-based collaborative filtering recommendation model. In relation to model selection based on overall experimental results, the MLP model showed highest performance with weighted average precision, weighted average recall, and f1 score were 0.85 in the model (n=3,000, term=15). This study is meaningful in practice as it attempted to commercialize a deep learning-based recommendation system where influencer's promotion data is being accumulated, pactical personalized recommendation service is not yet commercially applied yet.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.3
• /
• pp.71-78
• /
• 2011

This paper propose image steganalysis scheme based on independence between parent and child subband on the multi-layer wavelet domain. The proposed method decompose cover and stego images into 12 subbands by applying 3-level Haar UWT(Undecimated Wavelet Transform), analyze statistical independency between parent and child subband. Because this independency is appeared more difference in stego image than in cover image, we can use it as feature to differenciate between cover and stego image. Therefore we extract 72D features by calculation first 3 order statistical moments from joint characteristic function between parent and child subband. Multi-layer perceptron(MLP) is applied as classifier to discriminate between cover and stego image. We test the performance of proposed scheme over various embedding rates by the LSB, SS, BSS embedding method. The proposed scheme outperforms the previous schemes in detection rate to existence of hidden message as well as exactness of discrimination.

• Advances in concrete construction
• /
• v.13 no.1
• /
• pp.11-23
• /
• 2022

Conventional concrete needs some improvement in the mechanical properties, which can be obtained by different admixtures. However, making concrete samples costume always time and money. In this paper, different types of hybrid algorithms are applied to develop predictive models for forecasting compressive strength (CS) of concretes containing metakaolin (MK) and fly ash (FA). In this regard, three different algorithms have been used, namely multilayer perceptron (MLP), radial basis function (RBF), and support vector machine (SVR), to predict CS of concretes by considering most influencers input variables. These algorithms integrated with the grey wolf optimization (GWO) algorithm to increase the model's accuracy in predicting (GWMLP, GWRBF, and GWSVR). The proposed MLP models were implemented and evaluated in three different layers, wherein each layer, GWO, fitted the best neuron number of the hidden layer. Correspondingly, the key parameters of the SVR model are identified using the GWO method. Also, the optimization algorithm determines the hidden neurons' number and the spread value to set the RBF structure. The results show that the developed models all provide accurate predictions of the CS of concrete incorporating MK and FA with R² larger than 0.9972 and 0.9976 in the learning and testing stage, respectively. Regarding GWMLP models, the GWMLP1 model outperforms other GWMLP networks. All in all, GWSVR has the worst performance with the lowest indices, while the highest score belongs to GWRBF.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.53 no.8
• /
• pp.551-560
• /
• 2004

In this paper, we propose a new architecture of Self-Organizing Fuzzy Polynomial Neural Networks (SOFPNN) that is based on a genetically optimized multilayer perceptron with fuzzy polynomial neurons (FPNs) and discuss its comprehensive design methodology involving mechanisms of genetic optimization, especially genetic algorithms (GAs). The proposed SOFPNN gives rise to a structurally optimized structure and comes with a substantial level of flexibility in comparison to the one we encounter in conventional SOFPNNs. The design procedure applied in the construction of each layer of a SOFPNN deals with its structural optimization involving the selection of preferred nodes (or FPNs) with specific local characteristics (such as the number of input variables, the order of the polynomial of the consequent part of fuzzy rules, and a collection of the specific subset of input variables) and addresses specific aspects of parametric optimization. Through the consecutive process of such structural and parametric optimization, an optimized and flexible fuzzy neural network is generated in a dynamic fashion. To evaluate the performance of the genetically optimized SOFPNN, the model is experimented with using two time series data(gas furnace and chaotic time series), A comparative analysis reveals that the proposed SOFPNN exhibits higher accuracy and superb predictive capability in comparison to some previous models available in the literatures.