Facial Expression Recognition of Industrial Internet of Things by Parallel Neural Networks Combining Texture Features

Xi, Zhenghao; Niu, Yuhui; Chen, Jieyu; Kan, Xiu; Liu, Huaping

doi:10.1109/tii.2020.3007629

Cited by 49 publications

(17 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zeng et al [14] established a Deep Sparse Auto Encoder to learn robustness and distinguishability features from data to achieve high-precision recognition of facial expressions. Niu et al [15] proposed a parallel neural network that combines texture features in facial expression recognition, and Uddin et al [16] introduced a dynamic texture descriptor based on the Local Directional Structural Pattern from Three Orthogonal Planes, which can generate a stable description of facial dynamics. In order to solve the facial expression recognition problem of distinguishing highly similar expressions from easy to difficult under the influence of the class imbalance, Wang et al [17] proposed the Adaptive supervised objective (AdaReg) loss and Coarse-Fine (C-F) labelling strategy.…”

Section: Traditional Facial Expression Recognitionmentioning

confidence: 99%

See 1 more Smart Citation

Research on image sentiment analysis technology based on sparse representation

Jin

et al. 2022

CAAI Trans on Intel Tech

View full text Add to dashboard Cite

Many methods based on deep learning have achieved amazing results in image sentiment analysis. However, these existing methods usually pursue high accuracy, ignoring the effect on model training efficiency. Considering that when faced with large‐scale sentiment analysis tasks, the high accuracy rate often requires long experimental time. In view of the weakness, a method that can greatly improve experimental efficiency with only small fluctuations in model accuracy is proposed, and singular value decomposition (SVD) is used to find the sparse feature of the image, which are sparse vectors with strong discriminativeness and effectively reduce redundant information; The authors propose the Fast Dictionary Learning algorithm (FDL), which can combine neural network with sparse representation. This method is based on K‐Singular Value Decomposition, and through iteration, it can effectively reduce the calculation time and greatly improve the training efficiency in the case of small fluctuation of accuracy. Moreover, the effectiveness of the proposed method is evaluated on the FER2013 dataset. By adding singular value decomposition, the accuracy of the test suite increased by 0.53%, and the total experiment time was shortened by 8.2%; Fast Dictionary Learning shortened the total experiment time by 36.3%.

show abstract

Section: Traditional Facial Expression Recognitionmentioning

confidence: 99%

“…Niu et al. [15] proposed a parallel neural network that combines texture features in facial expression recognition, and Uddin et al. [16] introduced a dynamic texture descriptor based on the Local Directional Structural Pattern from Three Orthogonal Planes, which can generate a stable description of facial dynamics.…”

Section: Related Workmentioning

confidence: 99%

Research on image sentiment analysis technology based on sparse representation

Jin

et al. 2022

CAAI Trans on Intel Tech

View full text Add to dashboard Cite

show abstract

“…Last but not the least, Hao et al [ 34 ] worked on a similar approach as well but using DenseNet [ 35 ], a novel network architecture. Although both ResNet [ 36 ] and DenseNet [ 37 ] are famous for their skip connection approach, the latter achieved better performance via feature concatenate, whereas the former adopted feature addition.…”

Section: Introductionmentioning

confidence: 99%

Detection of Myocardial Infarction Using ECG and Multi-Scale Feature Concatenate

Jian

Ger

Lai

et al. 2021

Sensors

View full text Add to dashboard Cite

Diverse computer-aided diagnosis systems based on convolutional neural networks were applied to automate the detection of myocardial infarction (MI) found in electrocardiogram (ECG) for early diagnosis and prevention. However, issues, particularly overfitting and underfitting, were not being taken into account. In other words, it is unclear whether the network structure is too simple or complex. Toward this end, the proposed models were developed by starting with the simplest structure: a multi-lead features-concatenate narrow network (N-Net) in which only two convolutional layers were included in each lead branch. Additionally, multi-scale features-concatenate networks (MSN-Net) were also implemented where larger features were being extracted through pooling the signals. The best structure was obtained via tuning both the number of filters in the convolutional layers and the number of inputting signal scales. As a result, the N-Net reached a 95.76% accuracy in the MI detection task, whereas the MSN-Net reached an accuracy of 61.82% in the MI locating task. Both networks give a higher average accuracy and a significant difference of p < 0.001 evaluated by the U test compared with the state-of-the-art. The models are also smaller in size thus are suitable to fit in wearable devices for offline monitoring. In conclusion, testing throughout the simple and complex network structure is indispensable. However, the way of dealing with the class imbalance problem and the quality of the extracted features are yet to be discussed.

show abstract

“…As an important task in computer vision, multiple object tracking (MOT) has extremely important applications in intelligent surveillance [1,2], autonomous driving [3], medical diagnosis [4], and military vision guidance. Using intelligent technology to solve practical problems is becoming a trend [5,6,7]. However, MOT has to overcome more problems in tracking, such as occlusion and illumination.…”

Section: Introductionmentioning

confidence: 99%