文档介绍:基于粒子群算法的自适应滤波器研究与应用
摘 要
自适应滤波是数字信号处理技术的重要组成部分,对复杂信号的处理具有独特的功
能。在实际的数字信号系统中,叠加于信号的噪声、干扰往往不是单一的高斯噪声,而
线性滤波器所要求的中等程度噪声偏移,使线性滤波器对非高斯噪声的滤波性能下降,
为克服线性滤波器的缺点,往往采用非线性滤波器,为了适应自适应滤波实时化处理的
要求,文中引进了具有高度并行处理能力的人工神经网络来实现非线性自适应滤波。
粒子群优化算法是基于群智能方法的演化计算技术,相对其它进化算法,它收敛速
度快、规则简单、编程易于实现。针对粒子群优化算法易于陷入局部极值的缺点,本文
根据群体适应值标准差引入变异算子对算法进行改进。改进的算法摆脱了局部极值的束
缚,提高了非线性优化精度,同时保持了粒子群算法结构简单的特点。文中采用改进的
粒子群算法训练神经网络。
本文设计了基于粒子群算法的自适应滤波器,并将其应用到噪声抵消器中,对掺杂
有噪声的正弦信号进行滤波处理。改进的粒子群优化算法具有很强的处理能力和优化能
力,用它优化神经网络的权值,与传统的 BP 算法相比,达到了提高信噪比的目的同时,
还节省大量的学****和计算时间,进一步满足了自适应滤波实时处理的要求。经 Simulink
仿真实例证明基于粒子群算法的自适应噪声抵消器具有很强的噪声滤除能力。
关键词:自适应滤波,神经网络,粒子群优化算法,信噪比
I
The Study and Application of Adaptive Filter Based on Particle Swarm
Optimization Algorithm
ABSTRACT
Adaptive Filtering is the most important part of the digital signal processing technique,
having special function to the processing of complicated signal。In the actual digital signal
system, the noise which adds in the signal usually isn't single Gauss noise, but linear filter
requests medium degree noise excursion, making the filter function of the non- Gauss noise
descend. In order to overcome the weakness of the linear filtering, usually adopt non-linear
filtering. To meet the need of the adaptive filtering real-time, in this paper introduce a kind of
artificial neural networks with a high degree of parallel processing ability to achieve adaptive
non-linear filtering.
Particle swarm optimization (PSO) algorithm is a kind of evolution computing technique
based on swarm intelligence algorithms. To other evolutionary algorithms, its convergence’s
speed is fast, rules are simple, and it is easy to implement programming. In order to restrain
particles from tr