一种基于二阶梯度估计的自适应算法
An Adaptive Algorithm Based on the Second Derivative
-
摘要: 在自适应信号处理中得到广泛应用的LMS算法,对信号模型及特性有着极其严格的限 制,这些限制在很多实际情况中并不能保证得到满足.相对LMS算法,基于中心差的梯度估 计自适应算法,其适应面则要广泛得多.但是,该算法存在着收敛速度慢,所需采样点数多的 缺点.为此本文提出一种适应于平稳情况的新的估计算法,除首次估计需做采样外,在收敛过 程中无需再做采样.与传统的中心差算法相比,本文算法具有较快的收敛速度和较好的失调 性能.Abstract: The current LMS algorithms, widely applied to adaptive signal processing, have some rigorous restrictions on the model and its properties, which in practice may not be guaranteed. Adaptive algorithms based on central-differenc gradient estimation, on the other band, is better suited for many applications, but it requires extemsive samples and the convergence is slow. This paper proposes a new algorithm under stationary condition. The ,algroithm does not require sampling during convergence, converges fast, and has better property to cope with misadjustment.
-
Key words:
- Adaptive singal processing /
- adaptive filter /
- speech processing /
- neural networks
计量
- 文章访问数: 2899
- HTML全文浏览量: 94
- PDF下载量: 1021
- 被引次数: 0