Search

Feature fusion is an important part of building high-precision convolutional neural networks. In the field of image classification, though widely used in processing multiscale features of the same layer and short connections in the same receptive field, feature fusion is rarely used in long connection operations across receptive fields. In order to fuse the high- and low-level features of image classification, a feature fusion module SCFF (selective cross-layer feature fusion) for long connections is designed in this work. The SCFF can connect the long-distance feature maps in different receptive fields in a top-down order and apply the self-attention mechanism to fuse them two by two. The final fusion result is used as the input of the classifier. In order to verify the effectiveness of the model, the image classification experiment was done on a number of typical datasets. The experimental results prove that the model can fit the existing convolutional neural network well and effectively improve the classification accuracy of the convolutional network only at the cost of a small amount of calculation.

In Inconel 783 alloy, there are mainly two kinds of strengthening phases of β and γ’. In addition, it is found that there are other precipitated phases in the alloy’s β phase. By means of multiple testing methods, the types and functions of the precipitated phase in the β phase were determined. The results showed that the precipitated phases from β phase were nanoscale Nb and Ti carbides and a large number of white needle-like Laves phases. The former, due to its small size, which can improve the grain boundary slip resistance of the alloy, thus improves the shaping of the material. While the latter, due to its sharp shape and due to its large amount, destroys the original structure of the β phase and greatly reduces the room temperature plasticity of the alloy.

In-plant hot mix recycling technology has been widely used in maintenance and reconstruction due to its good recycling performance. However, the utilization rate of recycled asphalt pavement (RAP) is still only 10–25%, resulting in excessive accumulation and landfilling of RAP materials, which fails to promote the sustainable value of RAP in recycling. Therefore, this paper comprehensively explores the pavement performance of AC-20 ordinary asphalt recycled mixture and AC-13 modified asphalt recycled mixture with different proportions of RAP. The pavement performance test includes the dynamic modulus, rutting, dynamic creep, semicircular bending, and freeze-thaw splitting tests. The results show that, with the increase of RAP content, the mechanical properties and high-temperature properties of the two types of recycled asphalt mixtures are improved, the low temperature properties and moisture susceptibility properties are enhanced first and then weakened, and the best performance is reached when RAP content is 20 and 40%, respectively. According to the entire pavement performance test results, the influence of RAP content on AC-13 recycled modified asphalt mixtures is relatively smaller than that of AC-20 normal recycled asphalt mixtures. The addition of rejuvenator can improve the low-temperature performance and moisture susceptibility of recycled asphalt mixtures to a certain extent, especially for AC-20 ordinary recycled asphalt mixture. Consequently, it is suggested to use AC-13 recycled modified asphalt mixture in the upper layer of road; the RAP content can reach 20%, and rejuvenator can be used to improve its moisture susceptibility and low-temperature performance. AC-20 ordinary recycled asphalt mixture can be used in the middle surface layer, and the RAP content can reach 40%, or using AC-13 recycled modified asphalt mixture, the RAP content can reach 50%.

Recently, trimming Soft-output Viterbi algorithm (T-SOVA) has been proposed to reduce the complexity of SOVA for Turbo codes. In its first stage, a dynamic algorithm, lazy Viterbi algorithm, is used to indicate the minimal metric differences which brings obstacle on hardware implementation. This paper proposes a Viterbi algorithm (VA) based T-SOVA to facilitate hardware implementation. In the first stage of our scheme, a modified VA with regular structure is used to find the maximum likelihood (ML) path and calculate the metric differences. Further, local sorting is introduced to trim the metric differences, which reduces the complexity of trimming operation. Simulation results and complexity analysis show that VA based T-SOVA performs as well as lazy VA based T-SOVA and is easier to be applied to hardware implementation.

Iterative decoding of multilevel codes proposed in ITU-R BS.1514 and turbo code is explained clearly in this paper. Simulations made for a digital AM broadcasting (DAMB) system with these two channel coding algorithms respectively prove the advantage of turbo code both in performance and in complexity for the DAMB system.

High data transmission rates and high mobility give rise to time- and frequency-selectivity in wireless communication channels. This paper investigated time and frequency doubly selective channel estimation using pilot tones among Multi-Input Multi-Output (MIMO) orthogonal frequency division multiplexing (OFDM). Firstly, a complex exponential basis expansion channel model (BECM) was introduced to represent doubly selective channel during an OFDM symbol period; then, based on BECM, an effective MIMO-OFDM doubly selective channel estimation method was presented; finally, the optimality in designing pilot tones parameters was done according to the minimization of channel estimation MSE, mainly including the number, the placement and the structure of pilot tones. Simulation results show that the proposed estimation method has good performance in doubly selective channel scenarios and confirms the theoretical analysis findings.