ID	原文	译文
4823	该算法在每次迭代中，对天线功率约束集合进行最优选择，渐近地获得最优波束成形设计以平衡用户间的干扰，在满足 QoS 约束和天线功率约束下，考虑天线数量、用户数等因素的影响，实现异构大规模 MIMO 系统能效最优。	The algorithm selects the optimal antenna power constraint set through the optimization method, and the optimalbeamforming was asymptotically ob-tained to balance the interference among users to achieve the optimal energy ef-ficiency, considering the impact of the number of antennas and users with the constraints of the antenna power andQoS.
4824	鉴于回程功耗在大规模 MIMO 系统中的重要性，进一步考虑更符合实际应用的回程功耗系统模型，分析了回程功耗对系统性能的影响。	In view of the importance of backhaul in massive MIMO system, a backhaul power consumption model and theimpact of backhaul power consumption on system performance was analyzed.
4825	分析和仿真结果表明，提出的新算法具有较强的实用性和较大的性能提升，特别是当天线数较大时，算法性能接近最优，特别适合于下一代通信的大规模 MIMO 系统。	Analysis and simulation results show that the proposed algorithm has great improvement of the performance, especially when the number of antennas is large. The algorithm is close to optimal performance, especially suitable for massive MIMO system of next generationcommunication.
4826	基于复数支持向量回归机（CSVR）的框架，提出了一种针对复数信号的新的盲均衡算法，将多模算法的误差函数代入 CSVR 的惩罚项构造代价函数，利用广泛线性估计建立回归关系，并采用迭代重加权最小二乘方法确定均衡器系数。	A new blind equalization algorithm for complex valued signals was proposed based on the framework of complexsupport vector regression(CSVR). In the proposed algorithm, the error function of multi-modulus algorithm (MMA) wassubstituted into CSVR to construct the cost function, and the regression relationship was established by widely linear estima-tion, and the equalizer coefficients were determined by the iterative re-weighted least square (IRWLS) method.
4827	不同于支持向量回归机对复数信号的实数化处理方式，CSVR 利用 Wirtinger 微积分，将复数信号直接在复数再生核希尔伯特空间进行解析。	Different from spliting the complex valued signals into real valued signals used in support vector regression, the Wirtinger's calculus was used in complex support vector regression to analyze the complex signals directly in the complex regenerative kernel Hilbert space.
4828	仿真实验表明，针对 QPSK 调制信号，在线性信道和非线性信道下，与基于 SVR 的盲均衡算法相比，通过选取合适的核函数和迭代优化方法，所提算法的均衡性能显著提升。	Simulation experiments show that for QPSK modulated signals, compared with the blind equalization algo-rithm based on support vector regression, the equalization performance of the proposed algorithm is significantly improvedin linear channel and nonlinear channel by choosing appropriate kernel function and iterative optimization method.
4829	针对双流网络结构中浅层网络和一般深度模型学习空间信息和时间信息的不足，提出将压缩激励残差网络用于空间流和时间流的动作识别，	Aiming at the shortcomings of shallow networks and general deep models in two-stream network structure,which could not effectively learn spatial and temporal information, a squeeze-and-excitation residual network was pro-posed for action recognition with a spatial stream and a temporal stream.
4830	同时将恒等映射核作为时间滤波器注入网络中捕获长期时间依赖性。	Meanwhile, the long-term temporal dependence was captured by injecting the identity mapping kernel into the network as a temporal filter.
4831	为了进一步加强压缩激励残差网络的空间信息和时间信息之间的交互，采用时空特征相乘融合，并研究空间流和时间流乘法融合方式、次数以及位置对识别性能的影响。	Spatio temporal feature mul-tiplication fusion was used to further enhance the interaction between spatial information and temporal information ofsqueeze-and-excitation residual networks. Simultaneously, the influence of spatial-temporal stream multiplication fusionmethods, times and locations on the performance of action recognition was studied.
4832	鉴于单个模型获得性能的局限性，提出了 3 种不同的策略生成多个模型，并使用直接平均与加权平均集成以得到最终识别结果。	Given the limitations of performance achieved by a single model, three different strategies were proposed to generate multiple models, and the final recognition result was obtained by integrating these models through averaging and weighted averaging.