机械工程学报

EEG emotion recognition based on linear kernel PCA and XGBoost

Dong Yindong, Ren Fuji, Li Chunbin

2021, 48(2): 200013. doi: 10.12086/oee.2021.200013

[Abstract](228) [FullText HTML] (145)

Abstract:
The principal component analysis of linear kernel and XGBoost models are introduced to design electroencephalogram (EEG) classification algorithm of four emotional states under continuous audio-visual stimulation. In order to reflect universality, the traditional power spectral density (PSD) is used as the feature of EEG signal, and the feature importance measure under the weight index is obtained with XGBoost learning. Then linear kernel principal component analysis is used to process the threshold selected features and send them to XGBoost model for recognition. According to the experimental analysis, gamma-band plays a more important role than other bands in XGBoost model recognition; in addition, for distribution on channels, the central, parietal, and right occipital regions play a more important role than other brain regions. The recognition accuracy of this algorithm is 78.4% and 92.6% respectively under the two recognition schemes of subjects all participation (SAP) and subject single dependent (SSD). Compared with other literature, this algorithm has made a great improvement. The scheme proposed is helpful to improve the recognition performance of brain-computer emotion system under audio-visual stimulation.

RGB-D object recognition algorithm based on improved double stream convolution recursive neural network

Li Xun, Li Linpeng, Alexander Lazovik, Wang Wenjie, Wang Xiaohua

2021, 48(2): 200069. doi: 10.12086/oee.2021.200069

[Abstract](217) [FullText HTML] (144)

Abstract:
An algorithm (Re-CRNN) of image processing is proposed using RGB-D object recognition, which is improved based on a double stream convolutional recursive neural network, in order to improve the accuracy of object recognition. Re-CRNN combines RGB image with depth optical information, the double stream convolutional neural network (CNN) is improved based on the idea of residual learning as follows: top-level feature fusion unit is added into the network, the representation of federation feature is learning in RGB images and depth images and the high-level features are integrated in across channels of the extracted RGB images and depth images information, after that, the probability distribution was generated by Softmax. Finally, the experiment was carried out on the standard RGB-D data set. The experimental results show that the accuracy was 94.1% using Re-CRNN algorithm for the RGB-D object recognition, which was significantly improved compared with the existing image-based object recognition methods.

Research on multi-feature human pose model recognition based on one-shot learning

Li Guoyou, Li Chenguang, Wang Weijiang, Yang Mengqi, Hang Bingpeng

2021, 48(2): 200099. doi: 10.12086/oee.2021.200099

[Abstract](188) [FullText HTML] (190)

Abstract:
With the development of human-computer interaction, virtual reality, and other related fields, human posture recognition has become a hot research topic. Since the human body belongs to a non-rigid model and has time-varying characteristics, the accuracy and robustness of recognition are not ideal. Based on the KinectV2 somatosensory camera to collect skeletal information, this paper proposes a one-shot learning model matching method based on human body angle and distance characteristics. First, feature extraction is performed on the collected bone information, and the joint point vector angle and joint point displacement are calculated and a threshold is set. Secondly, the pose to be measured is matched with the template pose, and the recognition is successful if the threshold limit is met. Experimental results show that the method can detect and recognize human poses within the defined threshold in real-time, which improves the accuracy and robustness of recognition.

Polarization regulation characteristics of reflected waves at the interface of double topological insulators

Zhai Zhizhu, Wang Mingjun, Wu Biyuan

2021, 48(2): 200102. doi: 10.12086/oee.2021.200102

[Abstract](204) [FullText HTML] (174)

Abstract:
The property of reflective polarization on the surface of two 3D strong topological insulators was studied, obtaining the generalized necessary conditions required for the complete polarization conversion of linearly polarized light. By analyzing the reflectivity, cross reflectivity, and polarization conversion ratios of the interface of two topological insulators, we found that the model can realize the complete polarization conversion by using the existing topological insulator material, breaking through the limitation that the complete conversion requires a new small dielectric constant topological insulator material. The process can be verified by the Kerr rotation angle. Finally, we show the design method of polarization conversion devices to realize super strong angular stability. The polarization control capability of topological insulators can also be verified by Kerr effect. This work provides a theoretical basis for the application of topological insulators in polarized devices.

Design and simulation of bionic compound eye with electrowetting liquid lens

Zhao Rui, Peng Chao, Zhang Kai, Kong Meimei, Chen Tao, Guan Jianfei, Liang Zhongcheng

2021, 48(2): 200120. doi: 10.12086/oee.2021.200120

[Abstract](374) [FullText HTML] (190)

Abstract:
To solve the problem that the bionic compound eye system can't zoom adaptively, a zoomable bionic compound eye system based on electrowetting-on-dielectric liquid lens cambered array is proposed. The influence of the system structure on the imaging performance is analyzed, and the adaptive zoom capability of the system and the moving range of the corresponding image plane are calculated. The results show that the field of view angle increases with the increase of the curvature of the base. Compared with the non-uniform lens array, the uniform lens array can significantly reduce the defocus aberration of the system. Reducing the size of the lens unit can also decrease the defocus aberration of the edge lens. When the object distance or receiver position is changed, the defocus aberration of the system will be reduced by adjusting the focal length of the lens unit. The movable range of the system receiver is 1.9 mm~15 mm.

A B-spline based fast wavefront reconstruction algorithm

Chen Hao, Wei Ling, Li Ende, He Yi, Yang Jinsheng, Li Xiqi, Fan Xinlong, Yang Zeping, Zhang Yudong

2021, 48(2): 200160. doi: 10.12086/oee.2021.200160

[Abstract](230) [FullText HTML] (116)

Abstract:
Traditional schemes for Shack-Hartmann wavefront reconstruction can be classified into zonal and modal methods. The zonal methods are good at reconstructing the local details of the wavefront, but are sensitive to the noise in the slope data. The modal methods are much more robust to the noise, but they have limited capability of recovering the local details of the wavefront. In this paper, a B-spline based fast wavefront reconstruction algorithm in which the wavefront is expanded to the linear combination of bi-variable B-spline curved surfaces is proposed. Then, a method based on successive over relaxation (SOR) algorithm is proposed to fast reconstruct the wavefront. Experimental results show that the proposed algorithm can recover the local details of the wavefront as good as the zonal methods, while is much more robust to the slope noise.

A weakly supervised learning method for vehicle identification code detection and recognition

Cao Zhi, Shang Lidan, Yin Dong

2021, 48(2): 200170. doi: 10.12086/oee.2021.200170

[Abstract](185) [FullText HTML] (172)

Abstract:
The vehicle identification code (VIN) is of great significance to the annual vehicle inspection. However, due to the lack of character-level annotations, it is impossible to perform the single-character style check on the VIN. To solve this problem, a single-character detection and recognition framework for VIN is designed and a weakly supervised learning algorithm without character-level annotation is proposed for this framework. Firstly, the feature information of each level of VGG16-BN is fused to obtain a fusion feature map with single-character position information and semantic information. Secondly, a network structure for both the character detection branch and the character recognition branch is designed to extract the position and semantic information of a single character in the fusion feature map. Finally, using the text length and single-character category information, the proposed framework is weakly supervised on the vehicle identification code data set without character-level annotations. On the VIN test set, experimental results show that the proposed method realizes the Hmean score of 0.964 and a single-character detection and recognition accuracy rate of 95.7%, showing high practicability.

Target tracking algorithm based on YOLOv3 and ASMS

Lv Chen, Cheng Deqiang, Kou Qiqi, Zhuang Huandong, Li Haixiang

2021, 48(2): 200175. doi: 10.12086/oee.2021.200175

[Abstract](274) [FullText HTML] (235)

Abstract:
In order to solve the problem of loss when the target encounters occlusion or the speed is too fast during the automatic tracking process, a target tracking algorithm based on YOLOv3 and ASMS is proposed. Firstly, the target is detected by the YOLOv3 algorithm and the initial target area to be tracked is determined. After that, the ASMS algorithm is used for tracking. The tracking effect of the target is detected and judged in real time. Repositioning is achieved by quadratic fitting positioning and the YOLOv3 algorithm when the target is lost. Finally, in order to further improve the efficiency of the algorithm, the incremental pruning method is used to compress the algorithm model. Compared with the mainstream algorithms, experimental results show that the proposed algorithm can solve the lost problem when the tracking target is occluded, improving the accuracy of target detection and tracking. It also has advantages of low computational complexity, time-consuming, and high real-time performance.

A weakly supervised learning method for vehicle identification code detection and recognition

Cao Zhi, Shang Lidan, Yin Dong

2021, 48(2): 200270. doi: 10.12086/oee.2021.200270

[Abstract](271) [FullText HTML] (165)

Abstract:
The vehicle identification code (VIN) is of great significance to the annual vehicle inspection. However, due to the lack of character-level annotations, it is impossible to perform the single-character style check on the VIN. To solve this problem, a single-character detection and recognition framework for VIN is designed and a weakly supervised learning algorithm without character-level annotation is proposed for this framework. Firstly, the feature information of each level of VGG16-BN is fused to obtain a fusion feature map with single-character position information and semantic information. Secondly, a network structure for both the character detection branch and the character recognition branch is designed to extract the position and semantic information of a single character in the fusion feature map. Finally, using the text length and single-character category information, the proposed framework is weakly supervised on the vehicle identification code data set without character-level annotations. On the VIN test set, experimental results show that the proposed method realizes the Hmean score of 0.964 and a single-character detection and recognition accuracy rate of 95.7%, showing high practicability.

Research progress of electromagnetic properties of tunable chiral metasurfaces

Wang Jinjin, Zhu Qiuhao, Dong Jianfeng

2021, 48(2): 200218. doi: 10.12086/oee.2021.200218

[Abstract](333) [FullText HTML] (370)

Abstract:
Chiral metasurfaces are ultra-thin metamaterials composed of planar chiral cell structures with specific electromagnetic responses. They have attracted great attention due to their singular ability to control electromagnetic waves at will. With tunable materials incorporated into the metasurfaces design, one can realize tunable/reconfigurable metadevices with functionalities controlled by external stimuli, opening a new platform to dynamically manipulate electromagnetic waves. In this paper, we introduce some theoretical foundations of the electromagnetic properties of tunable/reconfigurable chiral metasurfaces. When a linearly polarized light enters a tunable chiral metasurface, it can be decomposed into left-handed circularly polarized (LCP) wave and right-handed circularly polarized (RCP) wave. By changing the dielectric constant and magnetic permeability of the medium through the external environment, the metadevices can dynamically control the response characteristics to various polarized lights, especially circularly polarized lights such as refractive index, dichroism, optical rotation, asymmetric transmission, etc. According to the properties of negative refractive index, circular dichroism, optical rotation, and asymmetric transmission controlled by the tunable chiral metasurfaces, we review the latest research progress. Finally, we put forward our own opinions on the possible future development directions and existing challenges of the rapidly developing field of the tunable chiral metasurface.

Light People: Professor Chennupati Jagadish

Wang Hui

2021, 10(2): 1146-1149. doi: 10.1038/s41377-021-00533-6

[Abstract](99) [FullText HTML] (84)

Abstract:
In 2018, the Indian film "Starting Line" focused the public's attention on the issue of education in India. It depicted the length some Indian parents were willing to go to secure educational resources for their children, as well as the difficulties faced by those disadvantaged in society in their fight for equal educational opportunities. In reality, many brilliant young Indian talents have been able to study in Australia through a fund set up by Prof. Chennupati Jagadish, a Distinguished Professor of the Australian National University. Prof. Jagadish is a Fellow of the Australian Academy of Science and the Australian Academy of Technological Sciences and Engineering. In 2018 he was awarded a UNESCO Prize for his contribution to the development of nanoscience and nanotechnology. He holds many positions, and has won numerous awards. What started Prof. Jagadish on his scientific research career? How did he become the respected scientist he is today? What was his intention in setting up the educational fund for students from developing countries? What advice does he have for young researchers? Here are the answers from Prof. Jagadish.

Silicon/2D-material photodetectors: from near-infrared to mid-infrared

Liu Chaoyue, Guo Jingshu, Yu Laiwen, Li Jiang, Zhang Ming, Li Huan, Shi Yaocheng, Dai Daoxin

2021, 10(2): 1164-1184. doi: 10.1038/s41377-021-00551-4

[Abstract](276) [FullText HTML] (209)

Abstract:
Two-dimensional materials (2DMs) have been used widely in constructing photodetectors (PDs) because of their advantages in flexible integration and ultrabroad operation wavelength range. Specifically, 2DM PDs on silicon have attracted much attention because silicon microelectronics and silicon photonics have been developed successfully for many applications. 2DM PDs meet the imperious demand of silicon photonics on low-cost, high-performance, and broadband photodetection. In this work, a review is given for the recent progresses of Si/2DM PDs working in the wavelength band from near-infrared to mid-infrared, which are attractive for many applications. The operation mechanisms and the device configurations are summarized in the first part. The waveguide-integrated PDs and the surface-illuminated PDs are then reviewed in details, respectively. The discussion and outlook for 2DM PDs on silicon are finally given.

Quantum state preparation and its prospects in quantum machine learning

Zhao Jian, Chen Zhao-Yun, Zhuang Xi-Ning, Xue Cheng, Wu Yu-Chun, Guo Guo-Ping

2021, 70(2): 140307. doi: 10.7498/aps.70.20210958

[Abstract](615) [FullText HTML] (319)

Abstract:
The development of traditional classic computers relies on the transistor structure of microchips, which develops in accordance with Moore's Law. In the future, as the distance between transistors approaches to the physical limit of manufacturing process, the development of computation capability of classical computers will encounter a bottleneck. On the other hand, with the development of machine learning, the demand for computation capability of computer is growing rapidly, and the contradiction between computation capability and demand for computers is becoming increasingly prominent. As a new computing model, quantum computing is significantly faster than classical computing for some specific problems, so, sufficient computation capability for machine learning is expected. When using quantum computing to deal with machine learning tasks, the first basic problem is how to represent the classical data effectively in the quantum system. This problem is called the state preparation problem. In this paper, the relevant researches of state preparation are reviewed, various state preparation schemes proposed at present are introduced, the processes of realizing these schemes are described, and the complexities of these schemes are summarized and analyzed. Finally, some prospects of the research work in the direction of state preparation are also presented.

Submit

Search

Search Type

Author FieldMORE+

Download CenterMORE+

2021 No. 2

NewsMORE+

LinksMORE+