• 스마트미디어저널
• /
• 제9권1호
• /
• pp.23-29
• /
• 2020

This paper presents an approach for dynamic hand gesture recognition by using algorithm based on 3D Convolutional Neural Network (3D_CNN), which is later extended to 3D Residual Networks (3D_ResNet), and the neural network based key frame selection. Typically, 3D deep neural network is used to classify gestures from the input of image frames, randomly sampled from a video data. In this work, to improve the classification performance, we employ key frames which represent the overall video, as the input of the classification network. The key frames are extracted by SegNet instead of conventional clustering algorithms for video summarization (VSUMM) which require heavy computation. By using a deep neural network, key frame selection can be performed in a real-time system. Experiments are conducted using 3D convolutional kernels such as 3D_CNN, Inflated 3D_CNN (I3D) and 3D_ResNet for gesture classification. Our algorithm achieved up to 97.8% of classification accuracy on the Cambridge gesture dataset. The experimental results show that the proposed approach is efficient and outperforms existing methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권7호
• /
• pp.3433-3446
• /
• 2019

Aiming at the high complexity of traditional single-channel demodulation algorithm for PCMA signals, a new demodulation algorithm based on neural network is proposed to reduce the complexity of demodulation in the system of non-cooperative PCMA communication. The demodulation network is trained in this paper, which combines the preprocessing module and decision module. Firstly, the preprocessing module is used to estimate the initial parameters, and the auxiliary signals are obtained by using the information of frequency offset estimation. Then, the time-frequency characteristic data of auxiliary signals are obtained, which is taken as the input data of the neural network to be trained. Finally, the decision module is used to output the demodulated bit sequence. Compared with traditional single-channel demodulation algorithms, the proposed algorithm does not need to go through all the possible values of transmit symbol pairs, which greatly reduces the complexity of demodulation. The simulation results show that the trained neural network can greatly extract the time-frequency characteristics of PCMA signals. The performance of the proposed algorithm is similar to that of PSP algorithm, but the complexity of demodulation can be greatly reduced through the proposed algorithm.

• 한국가시화정보학회지
• /
• 제20권2호
• /
• pp.78-85
• /
• 2022

In these days, the rapid development in prediction technology using artificial intelligent is being applied in a variety of engineering fields. Especially, dimensionality reduction technologies such as autoencoder and convolutional neural network have enabled the classification and regression of high-dimensional data. In particular, pixel level prediction technology enables semantic segmentation (fine-grained classification), or physical value prediction for each pixel such as depth or surface normal estimation. In this study, the pressure distribution of the ship's surface was estimated at the pixel level based on the artificial neural network. First, a potential flow analysis was performed on the hull form data generated by transforming the baseline hull form data to construct 429 datasets for learning. Thereafter, a neural network with a U-shape structure was configured to learn the pressure value at the node position of the pretreated hull form. As a result, for the hull form included in training set, it was confirmed that the neural network can make a good prediction for pressure distribution. But in case of container ship, which is not included and have different characteristics, the network couldn't give a reasonable result.

• 한국음향학회지
• /
• 제39권1호
• /
• pp.24-31
• /
• 2020

본 논문에서는 음향 이벤트 검출을 위한 심층 신경망에 대한 연구를 진행하였다. 특히 약하게 표기된 데이터 및 표기되지 않은 훈련 데이터를 포함하는 약지도 문제에 대하여, 입력 오디오 파형으로부터 이벤트 검출 결과를 얻어내는 종단간 신경망을 구축하는 연구를 진행하였다. 본 연구에서 제안하는 시스템은 1차원 콘볼루션 신경망을 깊게 적층하는 구조를 기반으로 하였으며, 도약 연결 및 게이팅 메커니즘 등의 추가적인 구조를 통해 성능을 개선하였다. 또한 음향 구간 검출 및 후처리를 통하여 성능을 향상시켰으며, 약지도 데이터를 다루기 위하여 평균-교사 모델을 적용하여 학습하는 과정을 도입하였다. 본 연구에서 고안된 시스템을 Detection and Classification of Acoustic Scenes and Events(DCASE) 2019 Task 4 데이터를 이용하여 평가하였으며, 그 결과 약 54 %의 구간-기반 F₁-score 및 32%의 이벤트-기반 F₁-score를 얻을 수 있었다.

• Nuclear Engineering and Technology
• /
• 제54권12호
• /
• pp.4751-4758
• /
• 2022

Two-phase flow may almost exist in every branch of the energy industry. For the corresponding engineering design, it is very essential and crucial to monitor flow patterns and their transitions accurately. With the high-speed development and success of deep learning based on convolutional neural network (CNN), the study of flow pattern identification recently almost focused on this methodology. Additionally, the photographing technique has attractive implementation features as well, since it is normally considerably less expensive than other techniques. The development of such a two-phase flow pattern online monitoring system is the objective of this work, which seldom studied before. The ongoing preliminary engineering design (including hardware and software) of the system are introduced. The flow pattern identification method based on CNNs and transfer learning was discussed in detail. Several potential CNN candidates such as ALexNet, VggNet16 and ResNets were introduced and compared with each other based on a flow pattern dataset. According to the results, ResNet50 is the most promising CNN network for the system owing to its high precision, fast classification and strong robustness. This work can be a reference for the online monitoring system design in the energy system.

• 대한기계학회논문집
• /
• 제18권7호
• /
• pp.1881-1897
• /
• 1994

This paper presents a neural network-based computational scheme to generate the optimized robotic assembly sequence for an assembly product consisting of a number of parts. An assembly sequence is considered to be optimal when it meets a number of conditions : it must satisfy assembly constraints, keep the stability of in-process subassemblies, and minimize assembly cost. To derive such an optimal sequence, we propose a scheme using both the Hopfield neural network and the expert system. Based upon the inferred precedence constraints and the assembly costs from the expert system, we derive the evolution equation of the network. To illustrate the suitability of the proposed scheme, a case study is presented for industrial product of an electrical relay. The result is compared with that obtained from the expert system.

• 한국멀티미디어학회논문지
• /
• 제23권8호
• /
• pp.977-985
• /
• 2020

Aiming at the problem of complex feature extraction and low accuracy in human action recognition, this paper proposed a network structure combining batch normalization algorithm with GoogLeNet network model. Applying Batch Normalization idea in the field of image classification to action recognition field, it improved the algorithm by normalizing the network input training sample by mini-batch. For convolutional network, RGB image was the spatial input, and stacked optical flows was the temporal input. Then, it fused the spatio-temporal networks to get the final action recognition result. It trained and evaluated the architecture on the standard video actions benchmarks of UCF101 and HMDB51, which achieved the accuracy of 93.42% and 67.82%. The results show that the improved convolutional neural network has a significant improvement in improving the recognition rate and has obvious advantages in action recognition.

• ETRI Journal
• /
• 제39권6호
• /
• pp.841-850
• /
• 2017

We investigate neural network image reconstruction for magnetic particle imaging. The network performance strongly depends on the convolution effects of the spectrum input data. The larger convolution effect appearing at a relatively smaller nanoparticle size obstructs the network training. The trained single-layer network reveals the weighting matrix consisting of a basis vector in the form of Chebyshev polynomials of the second kind. The weighting matrix corresponds to an inverse system matrix, where an incoherency of basis vectors due to low convolution effects, as well as a nonlinear activation function, plays a key role in retrieving the matrix elements. Test images are well reconstructed through trained networks having an inverse kernel matrix. We also confirm that a multi-layer network with one hidden layer improves the performance. Based on the results, a neural network architecture overcoming the low incoherence of the inverse kernel through the classification property is expected to become a better tool for image reconstruction.

• 한국정보통신학회논문지
• /
• 제21권6호
• /
• pp.1143-1148
• /
• 2017

현재 인공지능과 딥 러닝이 사회적인 이슈로 떠오르고 있는 추세이며, 다양한 분야에 이 기술들을 응용하고 있다. 인공지능 분야의 여러 알고리즘들 중에서 각광받는 방법 중 하나는 Convolutional Neural Network이다. Convolutional Neural Network를 적은 양의 데이터에서 이용하거나, Layer의 구조가 복잡하지 않은 경우에는 학습시간이 길지 않아 속도에 크게 신경 쓰지 않아도 되지만, 학습 데이터의 크기가 크고, Layer의 구조가 복잡할수록 학습시간이 상당히 오래 걸린다. 이로 인해 GPU를 이용하여 병렬처리를 하는 방법을 많이 사용하는데, 본 논문에서는 CUDA를 이용한 Convolutional Neural Network를 구현하였으며, 비교에 사용한 Framework/Program들 보다 학습속도가 빨라지고 큰 데이터를 학습 시키는데 더욱 효율적으로 진행하도록 한다.

• 한국콘텐츠학회논문지
• /
• 제7권12호
• /
• pp.292-302
• /
• 2007

최근 고객 선호도에 맞는 정보 또는 상품을 예측하기 위한 연구들이 활발히 진행되고 있다. 고객의 만족도를 향상시키기 위해서 먼저 불필요한 정보들을 제거시켜야 하며 이러한 정보 필터링은 내용기반 필터링, 협업 필터링 등 여러 가지 기법을 통해 연구되고 있다. 본 논문에서는 기존 필터링 기법들의 문제점으로 나타나고 있는 희소성과 확장성을 해결하기 위해서 역전파 신경망을 이용하여 연구를 수행하였다. 신경망의 훈련 데이터는 설문조사를 통해 얻어진 데이터를 사용하였다. 최종적으로 설문조사를 통해 데이터를 수집하고 신경망 기반 추천시스템의 프로토 타입을 제안하였고 기존 정보필터링 기법의 문제점을 개선하였다.