排序方式: 共有5条查询结果,搜索用时 15 毫秒
1
1.
2.
在电子回声消除应用中,为提高自适应算法的收敛速度,提出一种改进的仿射投影算法及其快速实现形式.新算法利用回声路径的稀疏结构特征,通过收敛步长控制矩阵,按滤波器各系数幅值大小,等比例地为其指定相应收敛步长,以加快大系数收敛,最终达到加快滤波器整体收敛速度的目的.对新算法进行的统计学分析,为其快速收敛于目标系统的算法特性提供了理论依据.仿真实验表明与传统自适应算法相比,新算法能减小稳态失调并大幅提高收敛速度,其低计算复杂度亦保证了系统的实时性. 相似文献
3.
The reconstruction of background noise from an error signal of an adaptive filter is a key is-sue for developing Variable Step-Size Normalized Least Mean Square (VSS-NLMS) algorithm in the context of Echo Cancellation (EC). The core param-eter in this algorithm is the Background Noise Pow-er (BNP ); in the estimation of BNP, the power difference between the desired signal and the filter output, statistically equaling to the error signal pow-er, has been widely used in a rough manner. In this study, a precise BNP estimate is implemented by multiplying the rough estimate with a corrective factor, taking into consideration the fact that the er-ror signal consists of background noise and mis-alignment noise. This corrective factor is obtained by subtracting half of the latest VSS value from 1 after analyzing the ratio of BNP to the misalignment noise. Based on the precise BNP estimate, the PVSS-NLMS algorithm suitable for the EC system is eventually proposed. In practice, the proposed al-gorithm exhibits a significant advantage of easier controllability application, as prior knowledge of the EC environment can be neglected. The simula-tion results support the preciseness of the BNP esti-mation and the effectiveness of the proposed algo-rithm. 相似文献
4.
针对当前热轧带钢表面缺陷检测中存在精度低及复杂背景干扰等问题,提出一种基于坐标注意力(coordinate attention, CA)的CA-YOLOv5缺陷检测方法。主要对YOLOv5的输入端、外加模块和检测端3个方面进行改进:在输入端,采用随机拼接4张或9张图片的方法对训练数据进行增广,并利用遗传算法(genertic algorithm, GA)对网络超参数进行寻优,使得YOLOv5更适用于带钢缺陷检测;在主干网络和外加模块之间引入CA机制,加强网络对缺陷深层特征的提取能力;最后,在检测端,对每一检测分支进行解耦,将检测的分类和位置回归两类任务分开,提升网络对缺陷的检测能力。在NEU-DET热轧带钢表面缺陷数据集上进行了验证实验,实验结果证明,CA-YOLOv5的均值平均精度(mean average precision,mAP)达到84.36%,不仅较原YOLOv5算法提升6.68%,而且优于其他先进的检测算法。 相似文献
5.
利用目标系统的稀疏性,提出基于延时估计的局部迭代(Delay Estimate-Selective Partial Update, DE-SPU)算法以降低滤波器有效长度。新算法先利用移动窗积分获取目标系统的延时估计以确定活跃系数位置,然后每次迭代均以自适应算法更新全部活跃系数与循环更新一段非活跃系数。活跃系数获得较高的更新频率以提高系统收敛速度,非活跃系数获得较低的更新频率以保证系统的跟踪能力。新算法用一个滤波器完成对稀疏系统的延时估计与活跃系数辨识,可有效避免双滤波器结构的信息冗余。最后以回声消除为应用背景对新算法进行实验仿真,仿真结果验证了新算法的有效性。 相似文献
1