Due to the diversity of text expressions, sentiment classification algorithm based on semantic understanding is difficult establish a perfect dictionary and sentence matching template, which leads strong limitations algorithm. In particular, it has certain difficulties in student sentiments. Based this, this paper analyzes model by neural network uses group as an example explore application classification. Moreover, regularization method added loss function LSTM so that output at any time...

In the era of artificial intelligence, traditional teaching models can be replaced by intelligent models, thereby effectively improving efficiency ideological and political teaching. This paper proposes a multi-frame sliding window double-threshold clutter map CFAR algorithm analyzes its detection probability false alarm formula. Moreover, system based on intelligence improved machine learning is designed B/S model. addition, this article practical performance model combined with actual...

This paper introduced electrostatic-assisted laser-induced breakdown spectroscopy (LIBS) to enhance spectral intensity and improve the quantitative accuracy of trace metal elements (Cu, Al, Zn, Ca Na) in aqueous solutions.

Hand gesture recognition is getting more and important in the area of rehabilitation human machine interface (HMI). However, most current approaches are difficult to achieve practical application because an excess sensors. In this work, we proposed a method recognize six common hand gestures establish optimal relationship between muscle by utilizing only two channels surface electromyography (sEMG). We integrated approach process sEMG data including filtering, endpoint detection, feature...

In this paper, we mainly refers to the network structure of Alexnet, traditional convolution layer is modified Multi-layer Perceptron (MLP) enhance face image feature extraction, adding Max-Feature-Map (MFP) excitation function segmentation noise signal and information improve recognition accuracy. The Center Loss loss added reduce distance between elements in same class, which can better generalize its features misjudgment caused by classes. CASIA-Web data set used for training testing....

This paper presents the implementation of an autonomous aerial manipulation using a hexacopter equipped with two DOF robotic arm. The kinematic and dynamic models are developed by considering characteristics combined platform. A novel adaptive sliding mode controller is proposed for both position velocity control. By building SSD (Single Shot Detection) deep neural network based on learning, object detection solution developed. three dimensional coordinates target relative to multirotor...

Purpose This research aims to analyze the diagnostic contribution of different discriminative regions breast ultrasound image and develop a more effective diagnosis method taking advantage regions' complementarity. Methods First, original as inner region lesion, marginal zone posterior echo lesion were defined. The pretrained Inception‐V3 network was used these regions. Then, applied extract deep features other three images. Since there are many features, principal components analysis (PCA)...

A close-loop algorithm based on electromyography (EMG) state-space model and measurement equation is developed for the estimation of continuous joint movements to achieve active control a lower limb rehabilitation robot. While general Hill muscle estimates only torque from EMG signals in an open-loop form, we integrate forward dynamics movement into established model. proposed get measured value through inertial unit (IMU). Nonlinear extended Kalman filter used combine these two model,...

Situational detection in the traffic system is of great significance to management and even urban management. Traditional methods are generally based on roadside equipment monitoring roads, it difficult support large-scale fine-grained incident detection. In this study, we propose a method applied mobile edge, which detects incidents video captured by vehicle cameras, so as overcome limitations terminal perception. For swarm intelligence detection, an improved YOLOv5s object network, adding...

The meticulous task of soil region classification is fundamental to the effective management resources and development accurate systems. These systems are crucial for the...

This study aimed to predict and fit the nonlinear dynamic grip force of human upper limb using surface electromyographic (sEMG) signals. The research employed a time-series-based neural network, NARX, establish mapping relationship between signals forearm muscle groups force. Three-channel signal acquisition equipment sensor were used record data subjects under specific conditions. After preprocessing data, including outlier removal, wavelet denoising, baseline drift correction, NARX model...

We divide the recognition process into "object detection" and "behavior prediction". Firstly, all objects in image are detected, then detection results used as input of behavior part to predict interaction actions between objects. In feature extraction, we add extra parameters sampling point each convolution kernel give characteristic deformation, so that network has better adaptability complex scenes. target, attention mechanism is combined with ResNet network, structure changed from...

With the development of power system, users begin to use their own supply in order improve economy, but this also leads occurrence risk self-provided supply. The actual distribution network has few samples and it is difficult identify by using conventional deep learning methods. In achieve high accuracy identification with small samples, paper proposes a combination transfer learning, convolutional block attention module (CBAM), neural (CNN) an active network. Firstly, be able further...