An Electrocardiogram Signal Classification Algorithm Based on Improved Deep Residual Shrinkage Networks北大核心CSCD

心电信号分类是医疗保健领域的重要研究内容.针对大多数方法不能很好地降低样本数量少的类别漏诊率,以及降低预处理操作的复杂性问题,提出了一种基于改进深度残差收缩网络(IDRSN)的心电信号分类算法(即DRSL算法).首先,使用合成少数类过采样技术(SMOTE)扩充数量少的类别样本,从而解决了类不平衡问题;其次,利用改进深度残差收缩网络提取空间特征,其残差模块可以避免网络层加深造成的过拟合,压缩激励和软阈值化子网络可以提取重要局部特征并自动去除噪声;然后,通过长短期记忆网络(LSTM)提取时间特征;最后,利用全连接网络输出分类结果.在MIT-BIH心律失常数据集上的实验结果表明,该算法的分类性能优于IDRSN、DRSN、GAN+2DCNN、CNN+LSTM_ATTENTION、SE-CNN-LSTM分类算法.     

Physical treatment synergized with natural surfactant for improving gas–water interfacial behavior and foam characteristics of α-lactalbumin

The purpose of this study was to investigate effect of physical treatment (ultrasound, U/high pressure homogenization, H/combined treatment, UH or HU) and surfactant (Mogroside V, Mog) on air/water interface adsorption and foaming properties of α-lactalbumin (ALa). Firstly, the binding of Mog and all physical-treated ALa was a static quenching process. Mog had the greatest binding affinity for HU-ALa among all treated samples. U or H treatment could change surface hydrophobicity of ALa/Mog complex. Secondly, at the molar ratio (ALa:Mog) of 1:50, foaming ability (FA) of all ALa samples got the maximum. The sequence of FA in ALa and ALa/Mog complex was listed as follow: HU > U > H > UH. Moreover, foaming stability (FS) of HU-ALa was the highest, followed by H-ALa, U-ALa and UH-ALa. Meanwhile, low concentration Mog increased FS of ALa or UH-ALa, but it reduced FS of H-ALa, U-ALa and HU-ALa. Quartz crystal microbalance with dissipation monitoring (QCM-D) experiment indicated that ALa/Mog complex after U or H treatment was quickly absorbed at air/water interface, compared with the treated ALa, and HU-ALa/Mog had the largest frequency shift. In addition, HU-ALa had the thickest bubble membrane and the highest dissipation shift in all samples, indicating that the absorbed membrane thickness and viscoelasticity of samples was correlated with foam stability. Therefore, U and H treatment synergism with Mog was an effective approach to enhance foam properties of ALa, which indicated that HU-treated ALa/Mog complex could be viewed as the safe and efficient foaming agent applied in food processing.     

含水煤岩变形破坏过程电荷感应和微震信号特征试验研究

为了监测煤层注水减弱煤层的冲击倾向性和底板突水增加矿井冲击危险性的效果,利用电荷感应和微震同步监测系统,对平顶山矿井的原煤和砂岩开展单轴压缩下不同含水率煤岩的电荷感应和微震监测试验,获得了不同含水率原煤和砂岩变形破裂过程的电荷感应和微震信号特征。试验结果表明：随着含水率的增大,煤岩峰值应力减小,冲击倾向性减弱,经过浸泡饱和后煤样的峰值应力和砂岩的峰值应力分别降低了21.8%和31.1%;随着含水率的增大,煤样变形破裂过程中产生的电荷感应和微震信号事件数都减小,信号强度也降低,电荷感应与微震信号均方值幅值减小,而砂岩变形破裂过程中产生的电荷感应和微震信号事件数却增多,信号强度也增强;煤岩变形破裂过程产生的电荷感应信号受水的影响比微震信号受水的影响大。煤层注水后有效降低了矿井的冲击危险性,煤层变形破裂过程产生的电荷感应和微震信号变弱;矿井底板突水时提高了矿井冲击危险性,从而底板变形破裂过程产生的电荷感应和微震信号增强。     

使用快速傅里叶变换优化周期参数的EMD-FFT-SARIMA光伏发电预测模型

根据分布式能源工业园区的光伏电力单元特点,对园区光伏发电功率预测模型进行优化,为后续的调度策略提供数据支持。针对经验模式分解（EMD）与季节性差分自回归移动平均模型（SARIMA）相组合的EMD-SARIMA预测模型中,原始数据经过EMD分解得到的各固有本征模态函数（IMF）分量周期计算问题,提出加入快速傅里叶变换（FFT）的周期计算方法,建立EMD-FFT-SARIMA光伏发电功率预测模型。再将每个IMF对应的预测结果进行叠加重构得到最终的预测结果。通过预测结果的误差计算可以发现,加入FFT环节后均方根误差（RMSE）从120.6 MW下降到19.3 MW,平均绝对误差（MAE）从52.87 MW下降到12.3 MW。     

COSTAS波形的主动声呐近似相干处理方法*

为了提升探测处理增益,有效对抗水声信道的频率选择性衰落,本文提出了一种基于COSTAS波形的主动声呐新算法。本文首先分析了相干融合的原理,然后基于相近频率经历近似相干信道的假设,提出了将近似相干融合应用于声呐回波信号处理,具体实现方式为如下步骤：1.通过频域波束形成提升回波增益;2.通过归一化匹配滤波进行相关处理;3将匹配滤波后的不同频率的复数进行加权求和,加权系数由信噪比决定。经过仿真表明,本文提出的近似相干算法的接收机(ROC)曲线明显优于非相干处理;同时,湖试结果证明,近似相干方法比现有的非相干处理方法有更高的处理增益,增益提升5dB~9dB。