Action recognition based on a human skeleton is an extremely challenging research problem. The temporal information contained in the human skeleton is more difficult to extract than the spatial information. Many researchers focus on graph convolution networks and apply them to action recognition. In this study, an action recognition method based on a two-stream network called RNXt-GCN is proposed on the basis of the Spatial-Temporal Graph Convolutional Network (ST-GCN). The human skeleton is converted first into a spatial-temporal graph and a SkeleMotion image which are input into ST-GCN and ResNeXt, respectively, for performing the spatial-temporal convolution. The convolved features are then fused. The proposed method models the temporal information in action from the amplitude and direction of the action and addresses the shortcomings of isolated temporal information in the ST-GCN. The experiments are comprehensively performed on the four datasets: 1) UTD-MHAD, 2) Northwestern-UCLA, 3) NTU RGB-D 60, and 4) NTU RGB-D 120. The proposed model shows very competitive results compared with other models in our experiments. On the experiments of NTU RGB?+?D 120 dataset, our proposed model outperforms those of the state-of-the-art two-stream models.
The formation mechanism of surface texture for feed-direction ultrasonic vibration?assisted milling (UVAM) was investigated in this study by establishing a kinematic model and a pressing model of UVAM. The kinematic model showed that the cutter tip, which was supposed to be totally sharp, produced closed scratches by crossed trajectories. The variation trends of the interval for closed textures were of sine function. A comparative experiment was carried out by UVAM and conventional milling. A dividing line close to the X coordinate divides the surface feature into two parts. The pressing model showed that the tool minor cutting edge left clear traces with certain width because of the tool minor cutting edge angle. Scratches by tool minor cutting edge were intermittent and regularly varied when feed-direction vibration was introduced. All the surface feature changes are in the radial direction and the trajectory intersections shall always be the scratch grooves or ridges. The ratio between ultrasonic vibration frequency and spindle speed, tool radius, and the located cutter rotation angle affected the changing rule of scratches by tool minor cutting edge. The analytical models and the experimental results proved to each other reasonable. 相似文献
Although the photoacoustic effect is almost universally generated by radiation whose intensity is varied in time either by amplitude modulation of a continuous optical source or through the use of pulsed irradiation, it is possible to produce sound by movement of a continuous source in space. Here, the characteristics of sound production by movement of a light source in one dimension are discussed by solution to the wave equation for pressure. Solutions to the wave equation for the velocity potential, from which the acoustic pressure can be determined, are found using the D’Alembert integral and by Fourier transformation of the wave equation. The characteristics of the waveform generated by a Gaussian heat source moving uniformly in space are found to depend on the initial conditions for movement of the source. 相似文献
Transcatheter aortic heart valves (TAHVs) have been widely used for aortic valve replacements, with less trauma and lower clinical risk compared with traditional surgical heart valve replacements. In the present study, composites of poly(ethylene glycol) diacrylate (PEGDA) hydrogels and anisotropic high-shrinkage polyethylene terephthalate/polyamide6 (PET-PA6) fabric (PEGDA/PET-PA6) were fabricated as artificial heart valve leaflets. Dynamic mechanical analyses (DMA) indicated that PEGDA/PET-PA6 composites possessed anisotropic mechanical properties (i.e., storage moduli ~23.30 ± 1.36 MPa parallel to the aligned fabric fibers and ~9.68 ± 0.90 MPa perpendicular to the aligned fibers at 1 Hz) that were comparable to aortic valve leaflets. The PEGDA/PET-PA6 composites with smooth surfaces were highly hydrophilic (contact angle ~41.6° ± 3.8°) and had low-fouling properties without platelet adhesion, suggesting a low risk of thrombogenicity when they interacted with blood. Furthermore, transcatheter aortic heart valves were fabricated using nitinol self-expanding frames and PEGDA/PET-PA6 composites as artificial leaflets, which presented excellent hemodynamic performance with a large orifice area (1.75 cm2) and low regurgitation (3.41%), thus meeting the requirements of ISO 5840-3 standard. Therefore, PEGDA/PET-PA6 composites had suitable mechanical properties, good biocompatibility, and low-fouling properties, indicating that they might be used for TAHVs in the future. 相似文献
When the Transformer proposed by Google in 2017, it was first used for machine translation tasks and achieved the state of the art at that time. Although the current neural machine translation model can generate high quality translation results, there are still mistranslations and omissions in the translation of key information of long sentences. On the other hand, the most important part in traditional translation tasks is the translation of key information. In the translation results, as long as the key information is translated accurately and completely, even if other parts of the results are translated incorrect, the final translation results’ quality can still be guaranteed. In order to solve the problem of mistranslation and missed translation effectively, and improve the accuracy and completeness of long sentence translation in machine translation, this paper proposes a key information fused neural machine translation model based on Transformer. The model proposed in this paper extracts the keywords of the source language text separately as the input of the encoder. After the same encoding as the source language text, it is fused with the output of the source language text encoded by the encoder, then the key information is processed and input into the decoder. With incorporating keyword information from the source language sentence, the model’s performance in the task of translating long sentences is very reliable. In order to verify the effectiveness of the method of fusion of key information proposed in this paper, a series of experiments were carried out on the verification set. The experimental results show that the Bilingual Evaluation Understudy (BLEU) score of the model proposed in this paper on the Workshop on Machine Translation (WMT) 2017 test dataset is higher than the BLEU score of Transformer proposed by Google on the WMT2017 test dataset. The experimental results show the advantages of the model proposed in this paper. 相似文献