极端学习机FIR滤波器的设计

The FIR Filter Design by the Extreme Learning Machine

极端学习机作为单隐层前向神经网络的一种典型学习算法,学习速度快,泛化能力好,在函数逼近和模式分类领域都有广泛应用.本文把极端学习机的应用拓展到滤波器的设计当中,通过对数字信号处理中实例的计算机仿真,验证该算法设计的FIR滤波器具有较好的性能,能够取得理想的滤波效果,通带与阻带边界频率容易精确控制,且初始条件随机给定,算法速度快.

Extreme Learning Machine algorithm is a standard learning algorithm of single hidden feedforward neural networks characterized by great generalization capability and fast learning speed, and is widely applied in the field of function approximations and pattern recognition. This paper discusses the design of FIR filters by ELM algorithm, whose excellent performance is evaluated by computer simulations of some problems from the field of DSP （digital signal processing）. The results show that the filter can achieve the desired results, the resistance band frequency of border zone can be controlled precisely with high efficiency, and the algorithm is fast.