The current dialogue emotion recognition models often overlook the coherence features and discourse structure information in context modeling. Therefore, this paper proposes a dialogue emotion recognition model based on coherence features and discourse structure. Firstly, discourse coherence detection is conducted to eliminate weak or incoherent discourse information, and both local and global coherent information are obtained by constructing a coherence matrix. Secondly, a dialogue parser is utilized to establish discourse structure relations, and a directed acyclic graph is employed to model the discourse structure while conveying both discourse structure information and speaker information. Finally, through interactive attention, coherent information and discourse information are interactively integrated to generate emotional labels. This paper validates the proposed model using two public datasets, with results indicating that compared to existing models, the proposed model demonstrates certain improvements in performance indices.

A novel ±45° dual-polarized microstrip patch antenna based on “L”-shaped probe feeding is designed and simulated for optimization. Two orthogonal “L”-shaped probes with a height difference are used for mutual coupling feeding to achieve dual polarization, significantly increasing the antenna's channel capacity. The copper metal columns between the two dielectric substrates play the role of feeding and radiation. The microstrip radiation patches are connected through metalized vias, and symmetrical “τ”-shaped grooves designed with openings facing inward are adopted to increase the resonance frequency point, the polarization patch is designed in an “S”-shape to expand the bandwidth. Simulation results show that the S11 of the antenna is less than -10 dB within the frequency bands of 2.08～2.77 GHz & 3.66～5.36 GHz, the relative impedance bandwidth is 66.15%, the gain is not less than 6 dBi, the radiation efficiency is above 90%, the isolation between ports is greater than 10 dB, and the cross-polarization level is greater than 20 dB. Physical fabrication and actual measurements demonstrate good agreement between the measured and simulated results at port 1, while minor deviations at port 2 are attributed to fabrication and testing conditions but remain within acceptable limits. Compared with similar studies, the proposed antenna features wide bandwidth, compact structure, and ease of fabrication, making it suitable for C-band（3700～4200 MHz）and WLAN（2400～2484 MHz and 5150～5350 MHz）wireless transceiver communication systems.

To address the poor performance of underwater object detection caused by light attenuation and scattering, this paper proposes an enhanced underwater object detection framework based on YOLOv8, named ERMS-YOLOv8, aiming to improve detection accuracy. The backbone is replaced with an efficient vision transformer（EfficientViT）to strengthen feature extraction of underwater organisms and reduce information loss. The neck adopts a reparameterized generalized-directional feature pyramid network（RepGFPN）to enhance the fusion of high-level semantic and low-level spatial features, enabling richer feature representation. A mixed local channel attention for object detection（MLCA）is introduced to integrate channel, spatial, local, and global channel information, thereby boosting the model's representational capacity. Additionally, a scalable intersection over union loss（SIoU）is employed to improve boundary prediction accuracy. Experimental re sults demonstrate that the proposed method achieves mAP values of 83.9% on the UPRC2021 dataset and 84.4% on the DUO dataset, outperforming the original YOLOv8 and exhibiting superior performance in underwater object detection.

This study proposes a wood board recognition method that integrates laser speckle technology with deep learning. Conventional photography and laser speckle imaging were employed to capture wood board images before and after modification treatments under both normal lighting and adverse conditions（including darkness and defocusing）. A corresponding dataset was then constructed. Classification experiments were conducted using the ResNet34 deep learning model. The results show that the ResNet34 model achieves high recognition accuracy when classifying laser speckle datasets and maintains good performance even under adverse environmental conditions. Furthermore, by introducing a convolutional block attention module（CBAM）to optimize the ResNet34 convolutional neural network, the classification accuracy for laser speckle images reached 93.29%. The combination of laser speckle technology and deep learning provides a low-environmental-requirement, efficient, and promising approach for wood board classification.

This study focuses on the problem of fractional channel parameters affecting channel estimation performance in intelligent reflecting surface-orthogonal time frequency space（IRS-OTFS）communication systems. A channel estimation method for IRS-OTFS systems is proposed by leveraging the sparsity of OTFS channels in the delay-Doppler（DD）domain. First, the joint sparsity channel estimation problem among channel parameters is transformed into a sparse signal recovery problem. Next, the fast iterative shrinkage/thresholding algorithm（FISTA）is introduced to solve this problem. The inputoutput relationship of the IRS-OTFS communication system is then derived. To address the issue of manual parameter tuning in traditional FISTA, a network architecture based on the FISTA algorithm is proposed. This architecture unfolds the iterative process of the sparse signal recovery algorithm into a neural network. The network is designed to automatically learn the optimal hyperparameters and nonlinear functions within the algorithm. Theoretical analysis and simulation results demonstrate that, under the same channel transmission conditions, the proposed algorithm achieves lower estimation error com pared to the benchmark algorithm.