International Journal of Pattern Recognition and Artificial Intelligence

Human–computer interaction systems have been developed in large numbers and quickly applied to sports. Badminton is the best sport for applying robotics because it requires quick recognition and fast movement. For the development of badminton recognition and tracking systems, it is important to accurately identify badminton, venues, and opponents. In this paper, we designed and developed a badminton recognition and tracking system using two 2 000 000-pixel high-speed cameras. The badminton tracking system has a transmission speed of 250fps and the maximum speed of the badminton resonator is 300km/h. The system uses the camera link interface Camera Link to capture images of high-speed cameras and process all captured images in real time using different regions of interest settings. In order to improve accuracy, we propose a new method for judging the center point of badminton. We have proposed a detector that detects the four corner points of the field by using the contour information of the badminton court when the approximate position of the badminton court is known. We set the sensing area according to the approximate position of the badminton court and use the histogram in the sensing area to select the point closest to the contour. Specify the intersection of the line as the corner point of the badminton court. The proposed angle detector has a high detection rate. It is more than 10 times more accurate than traditional detectors. The moving badminton is detected by an elliptical detector. We propose a method to find the center of the correct ellipse from the four candidates by selecting the four candidate contours of the ellipse. Compared to conventional circular detectors and points on three-dimensional coordinates, the proposed elliptical detector reduces the error by about 3mm.

With the continuous improvement and development of cameras network, surveillance video has become the data source of the column stream, which greatly promotes the development of cross-camera person re-identification (Re-ID). However, supervised learning requires a lot of effort to manually label cross-cameras pairwise training data, which is lack of scalability and practical in actual video surveillance because there is a lack of well-labeled pairs of positive and negative samples under each camera. For addressing these negative effects, we set judgment conditions by using the association ranking method to self-discover positive and negative track-lets pairs of anchors with none of the pairwise ID labels, thereby defining a triplet loss. In order to optimize association loss for learning effective discriminative feature, the triplet loss adds adaptive weights according to the degree of easy-hard samples to generate an Adaptive Weighted Conditional Triplet Loss. Besides, for increasing the accuracy of self-discovering cross-camera anchors independently, which means successfully mine mutually best-matched track-lets and merge them under cross-camera, we use the top-rank from the intra-camera ranking list as a self-matched query sample which can double verify the matched-degree between top-rank. And eventually, we establish a new Association Loss and Self-Discovery Learning (ALSL) model with a complete end-to-end manner. We use three standard datasets, PRID2011, iLIDS-VID and MARS, to train the model and the experimental results prove that ALSL rank-1 is better than some superior video-based unsupervised person Re-ID methods.

Deep learning is a popular topic among machine learning researchers nowadays, with great strides being made in recent years to develop robust artificial neural networks for faster convergence to a reasonable accuracy. Network architecture and hyperparameters of the model are fundamental aspects of model convergence. One such important parameter is the initial values of weights, also known as weight initialization. In this paper, we perform two research tasks concerned with the weights of neural networks. First, we develop three novel weight initialization algorithms inspired by the neuroscientific construction of the mammalian brains and then test them on benchmark datasets against other algorithms to compare and assess their performance. We call these algorithms the lognormal weight initialization, modified lognormal weight initialization, and skewed weight initialization. We observe from our results that these initialization algorithms provide state-of-the-art results on all of the benchmark datasets. Second, we analyze the influence of training an artificial neural network on its weight distribution by measuring the correlation between the quantitative metrics of skewness and kurtosis against the model accuracy using linear regression for different weight initializations. Results indicate a positive correlation between network accuracy and skewness of the weight distribution but no affirmative relation between accuracy and kurtosis. This analysis provides further insight into understanding the inner mechanism of neural network training using the shape of weight distribution. Overall, the works in this paper are the first of their kind in incorporating neuroscientific knowledge into the domain of artificial neural network weights.

Defocus blur detection aims at separating regions on focus from out-of-focus for image processing. With today’s popularity of mobile phones with portrait mode, accurate defocus blur detection has received more and more attention. There are many challenges that we currently confront, such as blur boundaries of defocus regions, interference of messy backgrounds and identification of large flat regions. To address these issues, in this paper, we propose a new deep neural network with both global and local pathways for defocus blur detection. In global pathway, we locate the objects on focus by semantical search. In local pathway, we refine the predicted blur regions via multi-scale supervisions. In addition, the refined results in local pathway are fused with searching results in global pathway by a simple concatenation operation. The structure of our new network is developed in a feasible way and its function appears to be quite effective and efficient, which is suitable for the deployment on mobile devices. It takes about 0.2s per image on a regular personal laptop. Experiments on both CUHK dataset and our newly proposed Defocus400 dataset show that our model outperforms existing state-of-the-art methods.

Road detection from aerial images still is a challenging task since it is heavily influenced by spectral reflectance, shadows and occlusions. In order to increase the road detection accuracy, a proposed method for road detection by GAC model with edge feature extraction and segmentation is studied in this paper. First, edge feature can be extracted using the proposed gradient magnitude with Canny operator. Then, a reconstructed gradient map is applied in watershed transformation method, which is segmented for the next initial contour. Last, with the combination of edge feature and initial contour, the boundary stopping function is applied in the GAC model. The road boundary result can be accomplished finally. Experimental results show, by comparing with other methods in $F$-measure system, that the proposed method can achieve satisfying results.

The optimization of the features is vital to effectively detecting facial expressions. This research work has optimized the facial features by employing the improved quantum-inspired gravitation search algorithm (IQI-GSA). The improvement to the quantum-inspired gravitational search algorithm (QIGSA) is conducted to handle the local optima trapping. The QIGSA is the amalgamation of the quantum computing and gravitational search algorithm that owns the overall strong global search ability to handle the optimization problems in comparison with the gravitational search algorithm. In spite of global searching ability, the QIGSA can be trapped in local optima in the later iterations. This work has adapted the IQI-GSA approach to handle the local optima, stochastic characteristics and maintaining balance among the exploration and exploitation. The IQI-GSA is utilized for the optimized features selection from the set of extracted features using the LGBP (a hybrid approach of local binary patterns with the Gabor filter) method. The system performance is analyzed for the application of automated facial expressions recognition with the classification technique of deep convolutional neural network (DCNN). The extensive experimentation evaluation is conducted on the benchmark datasets of Japanese Female Facial Expression (JAFFE), Radboud Faces Database (RaFD) and Karolinska Directed Emotional Faces (KDEF). To determine the effectiveness of the proposed facial expression recognition system, the results are also evaluated for the feature optimization with GSA and QIGSA. The evaluation results clearly demonstrate the outperformed performance of the considered system with IQI-GSA in comparison with GSA, QIGSA and existing techniques available for the experimentation on utilized datasets.

Audio-based healthcare technologies are among the most significant applications of pattern recognition and Artificial Intelligence. Lately, a major chunk of the World population has been infected with serious respiratory diseases such as COVID-19. Early recognition of lung health abnormalities can facilitate early intervention, and decrease the mortality rate of the infected population. Research has shown that it is possible to automatically monitor lung health abnormalities through respiratory sounds. In this paper, we propose an approach that employs filter bank energy-based features and Random Forests to classify lung problem types from respiratory sounds. The adventitious sounds, crackles and wheezes appear distinct to the human ear. Moreover, different sounds are characterized by different frequency ranges that are dominant. The proposed approach attempts to distinguish the adventitious sounds (crackles and wheezes) by modeling the human auditory perception of these sounds. Specifically, we propose a respiratory sounds representation technique capable of modeling the dominant frequency range present in such sounds. On a publicly available dataset (ICBHI) of size 6898 cycles spanning over 5h, our results can be compared with the state-of-the-art results, in distinguishing two different types of adventitious sounds: crackles and wheezes.

Anesthesiology aims to make anesthesia safer and increase the precision of prognoses. Correct assessment of the anesthesia depth is crucial to its safety. At present, intraoperative electroencephalogram (EEG) monitoring is the primary mode of anesthesia depth monitoring and judgment. However, most clinical anesthesiologists rely on commercial anesthesia depth monitors to judge anesthesia depth, such as bispectral index (BIS) and patient state index (PSI). This may lack an understanding of associated changes in brain wave quantization. Therefore, this study conducts quantitative analyses of EEG signals during anesthesia induction. EEG signals are processed within specific time windows and extracted brainpower density spectrum arrays with different frequency bands, brain electrical signal spectra, source frequencies and other key indicators. Analysis and comparison of these indicators clarifies patterns of variation in EEG signals during early anesthesia induction. The spectral edge frequencies (SEFs) of EEG signals within different time windows can be modeled accurately, from which the specific time points of EEG signal changes are derived. Furthermore, the relationship between patient age and the effect of anesthetic drugs is preliminarily investigated by analyzing the SEF variations of different age groups. This study quantifies changes in the EEG signals of patients at the initial stage of anesthesia induction and drug-related effects are observed, which opens a way for further exploration of EEG changes in patients under general anesthesia.

Exploration algorithms based on the Boustrophedon path seldom consider the impacts of a robot turning at corners on the exploration time. This paper proposes the Forecast-Island and Bidding A*-Euclidean Selecting Boustrophedon Coordination (FIBA*ESBC) algorithm to calculate the turning time at corners in the overall exploration time and introduces a method to estimate the walking time in the Boustrophedon paths in order to determine the directions for path execution. Typically, in bidding-based exploration tasks, the cost is the Euclidean distance between the current position of the robot and the target point. When there is an obstacle between two points, the cost is set to infinity. Therefore, the selected target point is sometimes not optimal. The FIBA*ESBC algorithm is based on the exploration cost of a combination of the Euclidean distance and A* algorithm walking path, which can effectively solve this problem. Because the bidding is based on a greedy algorithm, the robot has a small unexplored island in the later exploration stage; therefore, full exploration is not possible or requires a long time with several repeated paths. The FIBA*ESBC algorithm prioritizes the exploration and estimation of hidden and existing unexplored islands. It can realize complete exploration and decrease the exploration time. Through simulation experiments conducted using Gazebo and RViz, the feasibility of the FIBA*ESBC algorithm is verified. Moreover, a simulation experiment is conducted in MATLAB for comparison with other algorithms. The analysis of the experimental data shows that the proposed algorithm has a relatively short exploration time.

Various UNet architecture-based image change detection algorithms promote the development of image change detection, but there are still some defects. First, under the encoder–decoder framework, the low-level features are extracted many times in multiple dimensions, which generates redundant information; second, the relationship between each feature layer is not modeled so sufficiently that it cannot produce the optimal feature differentiation representation. This paper proposes a remote image change detection algorithm based on the multi-feature self-attention fusion mechanism UNet network, abbreviated as MFSAF UNet (multi-feature self-attention fusion UNet). We attempt to add multi-feature self-attention mechanism between the encoder and decoder of UNet to obtain richer context dependence and overcome the two above-mentioned restrictions. Since the capacity of convolution-based UNet network is directly proportional to network depth, and a deeper convolutional network means more training parameters, so the convolution of each layer of UNet is replaced as a separated convolution, which makes the entire network to be lighter and the model’s execution efficiency is slightly better than the traditional convolution operation. In addition to these, another innovation point of this paper is using preference to control loss function and meet the demands for different accuracies and recall rates. The simulation test results verify the validity and robustness of this approach.

Distributed Virtual Reality (DVR) is a combination of network and virtual reality technology, it could facilitate to construct a uniformly shared Distributed Virtual Environment (DVE) by using network to connect geographically distributed multiplayers. This paper concentrates on the theoretical research and practical development about Multiplayer Virtual Intelligent System (MVIS), and the main contribution could be summarized as two points. (1) Based on the DVR technology, this paper presented some theoretical research on MVIS, including the classification of virtual entities, communication pattern of entities, and the behavioral consistency research. Furthermore, a Multiplayer Earliest Deadline First (MEDF) program was proposed in order to guarantee the consistency of entities. (2) A prototype algorithm experiment system, called Multiplayer Graph-algorithm Intelligent System (MGIS), was designed. MGIS not onlyefficiently solves many problems in traditional computer algorithm teaching, such as high-abstraction, difficulty to understand, and lack of interaction mechanism; but also extends the application of DVR to cultural tourism, because MGIS is developed on the 3D scene of Lushan Mountain, which is one of the notable tourist attractions in China, and was included in the UNESCO World Heritage list in 1996. What i’s more, MGIS illustrates the ability of expression, applicability and generality of the theoretical research about MVIS.

Over the decades, permanent magnet synchronous motor (PMSM) has been widely used in coal mine production. In this paper, an optimized neural network predictive controller (NNPC) of permanent magnet direct drive belt conveyor system (BCS) for mining based on reduced order model (ROM) is established. First, in order to establish the full order model of the permanent magnet direct drive BCS, CEMA is used for dynamic analysis, and the dynamic equation of the permanent magnet direct drive BCS is established. Second, the Proper Orthogonal Decomposition (POD) method is used to reduce the order in this paper. Finally, the NNPC of permanent magnet direct drive BCS based on the ROM is proposed. The simulation result shows that the order of BCS model is effectively reduced by the POD method. The NNPC based on the ROM has a good performance in the control of permanent magnet direct drive BCS, and the error between of the full order model and the ROM is 0.19m/s.

Aiming at the low precision problem of multi-cylinder cooperative propulsion control in different regions of shield propulsion hydraulic systems under conditions of large load changes, this paper proposes a tracking differentiator and self-adaptive nonlinear PID (TD-NPID) control method to improve the synchronous control characteristics of shield propulsion hydraulic systems. First, the working principles of shield propulsion hydraulic systems were analyzed, and a mathematical model and TD-NPID controller were developed. Then, a simulation model was developed in AMESim-MATLAB environment, and the synchronous dynamic performances of fuzzy PID control, conventional PID control, and TD-NPID control were compared and analyzed. The results demonstrated that the shield propulsion hydraulic system with TD-NPID control had better servo tracking ability and steady-state performance than the systems with fuzzy or conventional PID control, which verified the feasibility of the application of TD-NPID control for the synchronous control of shield propulsion hydraulic systems.

In order to analyze the feasibility of integrating Sunan music into children’s music teaching through research, a knowledge-based system for children’s music teaching strategies based on case-based reasoning is proposed. The system can collect sufficient data and information, the exact orientation of piano learning, and the exact choice of teaching modes by questionnaires. Research data show that the vast majority of children like music teaching, and they like 76% more than they do not like, and more parents who express an indifferent attitude to music are 84% more than those who feel important. This experimental study shows that most teachers hold a positive attitude toward the development and use of folk music education resources in kindergartens, and parents do not have enough knowledge about folk music. The infiltration of excellent local culture in children’s music activities can improve students’ cultural literacy, let the new generation understand the tradition and the classics, and let children understand the local culture of southern Jiangsu and inherit the excellent culture of the Chinese nation through the understanding, feeling, and understanding of the local culture of southern Jiangsu. This is the new idea of local cultural heritage in southern Jiangsu and also the responsibility of kindergarten education.