JWSAA: Joint weak saliency and attention aware for person re-identification

MHSA-Net: Multihead Self-Attention Network for Occluded Person Re-Identification

This article presents a novel person reidentification model, named multihead self-attention network (MHSA-Net), to prune unimportant information and capture key local information from person images. MHSA-Net contains two main novel components: multihead self-attention branch (MHSAB) and attention competition mechanism (ACM). The MHSAB adaptively captures key local person information and then produces effective diversity embeddings of an image for the person matching. The ACM further helps filter out attention noise and nonkey information. Through extensive ablation studies, we verified that the MHSAB and ACM both contribute to the performance improvement of the MHSA-Net. Our MHSA-Net achieves competitive performance in the standard and occluded person Re-ID tasks.

Distract Your Attention: Multi-Head Cross Attention Network for Facial Expression Recognition

This paper presents a novel facial expression recognition network, called Distract your Attention Network (DAN). Our method is based on two key observations in biological visual perception. Firstly, multiple facial expression classes share inherently similar underlying facial appearance, and their differences could be subtle. Secondly, facial expressions simultaneously exhibit themselves through multiple facial regions, and for recognition, a holistic approach by encoding high-order interactions among local features is required. To address these issues, this work proposes DAN with three key components: Feature Clustering Network (FCN), Multi-head Attention Network (MAN), and Attention Fusion Network (AFN). Specifically, FCN extracts robust features by adopting a large-margin learning objective to maximize class separability. In addition, MAN instantiates a number of attention heads to simultaneously attend to multiple facial areas and build attention maps on these regions. Further, AFN distracts these attentions to multiple locations before fusing the feature maps to a comprehensive one. Extensive experiments on three public datasets (including AffectNet, RAF-DB, and SFEW 2.0) verified that the proposed method consistently achieves state-of-the-art facial expression recognition performance. The DAN code is publicly available.

Research on the Guidance of Youth Labor Education Based on the “Combination of Education and Production Labor” Program Based on the Deep Learning Model

At present, there is a lack of research on Marx's idea of “combining education and productive labor” and its guiding significance for youth labor education, and no effective teaching model has been formed. In response to this problem, this study proposes a semi-supervised deep learning model based on u-wordMixup (SD-uwM). When there is a shortage of labeled samples, semi-supervised learning uses a large number of unlabeled samples to solve the problem of labeling bottlenecks. However, since the unlabeled samples and labeled samples come from different fields, there may be quality problems in the unlabeled samples, which makes the generalization ability of the model worse., resulting in a decrease in classification accuracy. The model uses the u-wordMixup method to perform data augmentation on unlabeled samples. Under the constraints of supervised cross-entropy and unsupervised consistency loss, it can improve the quality of unlabeled samples and reduce overfitting. The comparative experimental results on the AGNews, THUCNews, and 20Newsgroups data sets show that the proposed method can improve the generalization ability of the model and also effectively improve the time performance. The study found that the SD-uwM model uses the u-wordMixup method to enhance the unlabeled samples and combines the idea of the Mean Teacher model, which can significantly improve the text classification performance. The SD-uwM model can improve the generalization ability and time performance of the model, respectively, 86.4 ± 1.3 and 90.5 ± 1.3. Therefore, the use of SD-uwM in Marx's program is of great practical significance for the guidance process of youth labor education.

Efficient recognition of dynamic user emotions based on deep neural networks

The key issue at this stage is how to mine the large amount of valuable user sentiment information from the massive amount of web text and create a suitable dynamic user text sentiment analysis technique. Hence, this study offers a writing feature abstraction process based on ON-LSTM and attention mechanism to address the problem that syntactic information is ignored in emotional text feature extraction. The study found that the Att-ON-LSTM improved the micro-average F1 value by 2.27% and the macro-average F value by 1.7% compared to the Bi-LSTM model with the added attentivity mechanisms. It is demonstrated that it can perform better extraction of semantic information and hierarchical structure information in emotional text and obtain more comprehensive emotional text features. In addition, the ON-LSTM-LS, a sentiment analysis model based on ON-LSTM and tag semantics, is planned to address the problem that tag semantics is ignored in the process of text sentiment analysis. The experimental consequences exposed that the accuracy of the ON-LSTM and labeled semantic sentiment analysis model on the test set is improved by 0.78% with the addition of labeled word directions compared to the model Att-ON-LSTM without the addition of labeled semantic information. The macro-averaged F1 value improved by 1.04%, which indicates that the sentiment analysis process based on ON-LSTM and tag semantics can effectively perform the text sentiment analysis task and improve the sentiment classification effect to some extent. In conclusion, deep learning models for dynamic user sentiment analysis possess high application capabilities.

Analysis and Prediction of Subway Tunnel Surface Subsidence Based on Internet of Things Monitoring and BP Neural Network

With the acceleration of the urban development process and the rapid growth of China's population, the subway has become the first choice for people to travel, and the urban underground space has been continuously improved. The subway construction has become the focus of urban underground space development in the 21st century. During the construction of subway tunnels, the problem of surface settlement will inevitably be caused, and the problem of surface settlement will have a certain safety impact on the safe use of surface buildings. The impact of surface construction is predicted, so as to select the best construction technology and avoid the problem of surface subsidence to the greatest extent. On the basis of analyzing the principle of surface subsidence, this paper studies the optimal control strategy and process of subsidence in subway tunnel engineering. The research results of the article show the following. (1) The two sections of the pebble soil layer have basically the same subsidence trend. Among them, the first section has a larger settlement amplitude and both sides are steeper. The second section is mainly cobble clay soil. The pebble layer has good mechanical properties. If it can be well filled, its stability will be improved. The comparative analysis of the two sections shows that with the increase of the soil cover thickness, the maximum subsidence at the surface gradually decreases. The reason is that when the stratum loss is the same, the greater the soil cover thickness, the greater the settlement width. Sections 2 and 3 of a single silty clay have relatively close settlement laws, and the settlement changes on both sides of the tunnel are similar. (2) The surface subsidence caused by the excavation of the side hole accounts for more than 50% of the total surface subsidence, and the width of the settlement tank after the excavation of the side hole is increased by 8–10 meters compared with the excavation of the middle hole. (3) The prediction error of the BP neural network model proposed in this paper is the lowest among the four models, whether it is the prediction of the cumulative maximum surface subsidence or the location of the cumulative maximum surface subsidence, and the average relative error of the cumulative maximum surface subsidence is 3.27%, the root mean square error is 3.87, the average relative error of the location of the cumulative maximum surface subsidence is 7.96%, and the root mean square error is 21.06. In the prediction process of the cumulative maximum surface subsidence, the prediction error value of the Elman neural network is relatively large, and the GRNN generalized neural network and RBF neural network have no significant changes; in the process of predicting the position where the cumulative maximum surface subsidence occurs, the prediction error value of RBF neural network is maximum.

An information-rich sampling technique over spatio-temporal CNN for classification of human actions in videos

We propose a novel video sampling scheme for human action recognition in videos, using Gaussian Weighing Function. Traditionally in deep learning-based human activity recognition approaches, either a few random frames or every k ^{t h} frame of the video is considered for training the 3D CNN, where k is a small positive integer, like 4, 5, or 6. This kind of sampling reduces the volume of the input data, which speeds-up the training network and also avoids overfitting to some extent, thus enhancing the performance of the 3D CNN model. In the proposed video sampling technique, consecutive k frames of a video are aggregated into a single frame by computing a Gaussian-weighted summation of the k frames. The resulting frame preserves the information in a better way than the conventional approaches and experimentally shown to perform better. In this paper, a 3-Dimensional deep CNN is proposed to extract the spatio-temporal features and follows Long Short-Term Memory (LSTM) to recognize human actions. The proposed 3D CNN architecture is capable of handling the videos where the camera is placed at a distance from the performer. Experiments are performed with KTH, WEIZMANN, and CASIA-B Human Activity and Gait datasets, whereby it is shown to outperform state-of-the-art deep learning based techniques. We achieve 95.78%, 95.27%, and 95.27% over the KTH, WEIZMANN, and CASIA-B human action and gait recognition datasets, respectively.

Analysis of Psychological Changes and Intervention Mechanism of Elderly Groups Based on Deep Learning Analysis Technology

The elderly group is a unique social phenomenon in China. This study analyzes the typology of psychological changes in the elderly group based on the analysis of deep learning techniques and also combines crisis intervention theory to study the intervention strategies of social workers in different stages of emotional changes in the elderly group. A questionnaire survey of the elderly was conducted using the survey method, in which 10,948 valid questionnaires were screened from the Psychological Condition Self-check Scale and 11,040 valid questionnaires were screened from the Mental Health Survey Questionnaire for the Elderly. The degree of negative emotions of the elderly group in public emergencies was not related to age, but significantly correlated with age (p value < 0.05), and there was a tendency that the higher the age, the deeper the degree; in addition, elderly people of different professions also showed significant differences (p < 0.05); elderly people of different regions also showed significant differences (p < 0.05). In crisis intervention, social workers mainly provide services such as initial diagnosis, primary intervention, secondary intervention, and assessment for the caseworkers. The practical study found that social workers need to use strategies such as short-term focal solutions, avoiding guiding clients with their own values; crisis intervention programmes should be flexible, proactively helping clients to rebuild their support network, developing clients' self-solving skills, and implementing interprofessional teamwork and whole-person rehabilitation services.

5G Network Slicing: Methods to Support Blockchain and Reinforcement Learning

With the advent of the 5G era, due to the limited network resources and methods before, it cannot be guaranteed that all services can be carried out. In the 5G era, network services are not limited to mobile phones and computers but support the normal operation of equipment in all walks of life. There are more and more scenarios and more and more complex scenarios, and more convenient and fast methods are needed to assist network services. In order to better perform network offloading of the business, make the business more refined, and assist the better development of 5G network technology, this article proposes 5G network slicing: methods to support blockchain and reinforcement learning, aiming to improve the efficiency of network services. The research results of the article show the following: (1) In the model testing stage, the research results on the variation of the delay with the number of slices show that the delay increases with the increase of the number of slices, but the blockchain + reinforcement learning method has the lowest delay. The minimum delay can be maintained. When the number of slices is 3, the delay is 155 ms. (2) The comparison of the latency of different types of slices shows that the latency of 5G network slicing is lower than that of 4G, 3G, and 2G network slicing, and the minimum latency of 5G network slicing using blockchain and reinforcement learning is only 15 ms. (3) In the detection of system reliability, reliability decreases as the number of users increases because reliability is related to time delay. The greater the transmission delay, the lower the reliability. The reliability of supporting blockchain + reinforcement learning method is the highest, with a reliability of 0.95. (4) Through the resource utilization experiment of different slices, it can be known that the method of blockchain + reinforcement learning has the highest resource utilization. The resource utilization rate of the four slices under the blockchain + reinforcement learning method is all above 0.8 and the highest is 1. (5) Through the simulation test of the experiment, the results show that the average receiving throughput of video stream 1 is higher than that of video stream 2, IOT devices and mobile devices, and the average cumulative receiving throughput under the blockchain + reinforcement learning method. The highest is 1450 kbps. The average QOE of video stream 1 is higher than that of video stream 2, IOT devices and mobile devices, and the average QOE is the highest under the blockchain + reinforcement learning method, reaching 0.83.

A Bearing Fault Diagnosis Method Based on Wavelet Packet Transform and Convolutional Neural Network Optimized by Simulated Annealing Algorithm

Bearings are widely used in various electrical and mechanical equipment. As their core components, failures often have serious consequences. At present, most parameter adjustment methods are still manual adjustments of parameters. This adjustment method is easily affected by prior knowledge, easily falls into the local optimal solution, cannot obtain the global optimal solution, and requires a lot of resources. Therefore, this paper proposes a new method for bearing fault diagnosis based on wavelet packet transform and convolutional neural network optimized by a simulated annealing algorithm. Firstly, the original bearing vibration signal is extracted by wavelet packet transform to obtain the spectrogram, and then the obtained spectrogram is sent to the convolutional neural network for parameter adjustment, and finally the simulated annealing algorithm is used to adjust the parameters. To verify the effectiveness of the method, the bearing database of Case Western Reserve University is used for testing, and the traditional intelligent bearing fault diagnosis methods are compared. The results show that the new method for bearing fault diagnosis proposed in this paper has a better and more reliable diagnosis effect than the existing machine learning and deep learning methods.

Collaborative Research on Mouth Shape and Lyrics in Singing Practice Based on Image Processing

Image processing is a mainstream processing method. When people enjoy artists' singing videos, there will be a problem that the subtitles of the lyrics are out of sync with the singer's mouth shape. This problem needs to be solved using image processing technology, letting the computer realize lip-reading recognition function and correct the mouth shape and lyrics subtitles in the image according to the extracted lip-reading data, so that the mouth shape and lyrics in singing practice can be synchronized. Lip-reading information can effectively improve the accuracy of language cognition, save part of capital and manpower investment, and make viewers get a good audio-visual interactive experience. The results show the following: (1) After the UI test, the system user interface function design is reasonable and there is no bad BUG. We can find that the average processing time of each frame is 628 ms, the system performance evaluation is good, and the success rate can be as high as 98.80%. 0.36724 s is the average time for each step when the system processes the image. (2) The human image can basically identify the portrait area and lip area from various angles. (3) Compared with DCT and DWT, the recognition rate of the two cascade lip region feature extraction methods is improved by nearly 10%, and the feature vector dimension is reduced by nearly 65%. (4) Classify the mouth shape more finely and optimize the image of the tester's mouth shape to make the mouth shape closer to the standard mouth shape. (5) After systematic correction of mouth shape and subtitles, the success rate is higher than 90%. Finally, we can find that the running effect is good and the method has achieved high results, which can carry out the details of the next optimization work.

Research on Teaching Resource Recommendation Algorithm Based on Deep Learning and Cognitive Diagnosis

With the increasing abundance of network teaching resources, the recommendation technology based on network is becoming more and more mature. There are differences in the effect of recommendation, which leads to great differences in the effect of recommendation algorithms for teaching resources. The existing teaching resource recommendation algorithm either takes insufficient consideration of the students' personality characteristics, cannot well distinguish the students' users through the students' personality, and pushes the same teaching resources or considers the student user personality not sufficient and cannot well meet the individualized learning needs of students. Therefore, in view of the above problem, combining TDINA model by the user for the students to build cognitive diagnosis model, we put forward a model based on convolution (CUPMF) joint probability matrix decomposition method of teaching resources to recommend the method combined with the history of the students answer, cognitive ability, knowledge to master the situation, and forgetting effect factors. At the same time, CNN is used to deeply excavate the test question resources in the teaching resources, and the nonlinear transformation of the test question resources output by CNN is carried out to integrate them into the joint probability matrix decomposition model to predict students' performance on the resources. Finally, the students' knowledge mastery matrix obtained by TDINA model is combined to recommend corresponding teaching resources to students, so as to improve learning efficiency and help students improve their performance.

E-Commerce Picture Text Recognition Information System Based on Deep Learning

For the accuracy requirements of commodity image detection and classification, the FPN network is improved by DPFM ablation and RFM, so as to improve the detection accuracy of commodities by the network. At the same time, in view of the narrowing of channels in the application of traditional MWI-DenseNet network, a new GTNet network is proposed to improve the classification accuracy of commodities.The results show that at different levels of evaluation indexes, the dpFPN-Netv2 algorithm improved by DPFM + RFM fusion has higher target detection accuracy than RetinaNet-50 algorithm and other algorithms. And the detection time is 52 ms, which is significantly lower than 90 ms required for RetinaNet-50 detection. In terms of target recognition, compared with the traditional MWI-DenseNet neural network, the computation amount of the improved MWI DenseNet neural network is significantly reduced under different shunt ratios, and the recognition accuracy is significantly improved. The innovation of this study lies in improving the algorithm from the perspective of target detection and recognition, so as to change the previous improvement that only can be made in a single way.

Prediction Algorithm of Young Students' Physical Health Risk Factors Based on Deep Learning

Young people's physical and mental health is the foundation of society's overall development and the key to improving people's health quality. Middle school students' physical examinations and monitoring work are a surefire way to ensure their healthy development. Poor vision, dental caries, overweight and obesity, and high blood pressure are the most common adverse health outcomes of students caused by adolescent health risk behavior factors. Researchers have been concerned about the retinal fundus vascular system, which is the only internal vascular system that can be observed in a noninvasive state of the human body. Fundus images contain a wealth of disease-related information. Fundus images have been widely used in the field of medical auxiliary diagnosis because many important systemic diseases of the human body cause specific reactions in the fundus. Aiming to solve the problem of inseparable tiny blood vessels, this paper proposes a model of retinal vessel segmentation based on attention mechanisms. In light of the retinal arteriovenous division of discontinuous challenges, the topological structure of the constraint system along with overcoming the network and topology restrictions is monitored. Finally, simulation experiments were conducted on two publicly available datasets. The findings show that the proposed method is reliable, effective, and accurate in predicting physical health risk factors in adolescent students.

Wearable Device-Based Smart Football Athlete Health Prediction Algorithm Based on Recurrent Neural Networks

For football players who participate in sports, the word “health” is extremely important. Athletes cannot create their own value in competitive competitions without a strong foundation. Scholars have paid a lot of attention to athlete health this year, and many analysis methods have been proposed, but there have been few studies using neural networks. As a result, this article proposes a novel wearable device-based smart football player health prediction algorithm based on recurrent neural networks. To begin, this article employs wearable sensors to collect health data from football players. The time step data are then fed into a recurrent neural network to extract deep features, followed by the health prediction results. The collected football player health dataset is used in this paper to conduct experiments. The simulation results prove the reliability and superiority of the proposed algorithm. Furthermore, the algorithm presented in this paper can serve as a foundation for the football team's and coaches' scientific training plans.

Martial Arts Routine Training Method Based on Artificial Intelligence and Big Data of Lactate Measurement

As a traditional Chinese sport, competitive martial arts routines have a long history. The competition rules are the unified norms and standards formulated for sports competitions. They are a yardstick for referees to judge the technical level and competitive ability of athletes and an essential basis for coaches during training. In particular, the new rules increase the difficulty of martial arts routines training and score, improve the balance movement of various groups, highlight the action specifications, increase the proportion of the score, and strengthen the scoring measures for the performance level. Subsequently, this puts higher requirements for the exceptional technical level of routine athletes. Therefore, it is vital to formulate scientific martial arts systematic training methods. This paper considers the above problem and current popular artificial intelligence technology and constructs a neural network algorithm to solve it. In addition, since lactic acid is a good monitoring indicator of the training load intensity and effect of martial arts routine exercises, this article also considers extensive lactate measurement data to construct martial arts systematic training methods. Through simulations, our experimental verification and the obtained results demonstrate the effectiveness of the proposed algorithm.

Multi-angle head pose classification with masks based on color texture analysis and stack generalization

Head pose classification is an important part of the preprocessing process of face recognition, which can independently solve application problems related to multi-angle. But, due to the impact of the COVID-19 coronavirus pandemic, more and more people wear masks to protect themselves, which covering most areas of the face. This greatly affects the performance of head pose classification. Therefore, this article proposes a method to classify the head pose with wearing a mask. This method focuses on the information that is helpful for head pose classification. First, the H-channel image of the HSV color space is extracted through the conversion of the color space. Then use the line portrait to extract the contour lines of the face, and train the convolutional neural networks to extract features in combination with the grayscale image. Finally, stacked generalization technology is used to fuse the output of the three classifiers to obtain the final classification result. The results on the MAFA dataset show that compared with the current advanced algorithm, the accuracy of our method is 94.14% on the front, 86.58% on the more side, and 90.93% on the side, which has better performance.

Falling-Point Recognition and Scoring Algorithm in Table Tennis Using Dual-Channel Target Motion Detection

In table tennis, the ball has numerous characteristics of high speed, small size, and changeable trajectory. Due to these characteristics, the human eye often cannot accurately judge the ball's movement and position, leading to the problem of precise detection of the ball's falling point and movement tracking. In sports, the use of machine learning for locating and detecting the ball and the use of deep learning for reconstructing and displaying the ball's trajectories are considered futuristic technologies. Therefore, this paper proposes a novel algorithm for identifying and scoring points in table tennis based on dual-channel target motion detection. The proposed algorithm consists of multiple input channels to jointly learn different features of table tennis images. The original image is used as the input of the first channel, and then the Sobel operator is used to extract the first-order derivative feature of the original image, which is used as the input of the second channel. The table tennis feature information from the two channels is then fused and sent to the 3D neural network module. The fully connected layer is used to identify the table tennis ball's drop point, compare it with a standard drop point, calculate the error distance, and give a score. We also constructed a data set and conducted experiments. The experimental results show that the method in this paper is effective in sports.

An Improved Encoder-Decoder Network Based on Strip Pool Method Applied to Segmentation of Farmland Vacancy Field

In the research of green vegetation coverage in the field of remote sensing image segmentation, crop planting area is often obtained by semantic segmentation of images taken from high altitude. This method can be used to obtain the rate of cultivated land in a region (such as a country), but it does not reflect the real situation of a particular farmland. Therefore, this paper takes low-altitude images of farmland to build a dataset. After comparing several mainstream semantic segmentation algorithms, a new method that is more suitable for farmland vacancy segmentation is proposed. Additionally, the Strip Pooling module (SPM) and the Mixed Pooling module (MPM), with strip pooling as their core, are designed and fused into the semantic segmentation network structure to better extract the vacancy features. Considering the high cost of manual data annotation, this paper uses an improved ResNet network as the backbone of signal transmission, and meanwhile uses data augmentation to improve the performance and robustness of the model. As a result, the accuracy of the proposed method in the test set is 95.6%, mIoU is 77.6%, and the error rate is 7%. Compared to the existing model, the mIoU value is improved by nearly 4%, reaching the level of practical application.