论文部分内容阅读
ＦＩＲ滤波器的设计是数字信号处理中的一个重要问题，神经网络则在优化设计方面显示了其良好的应用前景，但目前大多数应用均限于实数神经网络的情形．本文针对ＦＩＲ数字滤波器的设计问题，构造了具有复数状态、输出、联接权和偏置输入的Ｈｏｐｆｉａｌｄ网络，给出了网络的计算能量函数，并证明了算法的收敛性．通过近似线性相位、非共轭对称频响ＦＩＲ滤波器的设计实例表明，本文的工作对推广复数神经网络的应用与寻求更为复杂的ＦＩＲ滤波器设计的新方法均作了有效的尝试．
The design of FIR filter is an important issue in digital signal processing. Neural network shows its good application prospect in optimization design. However, most applications are limited to the case of real neural network. In this paper, for the design of FIR digital filter, a Hopfiald network with complex states, outputs, coupling weights and bias inputs is constructed. The computational energy function of the network is given and the convergence of the algorithm is proved. Through the approximation of linear phase, the design example of FIR filter with non-conjugate symmetric frequency response shows that the work in this paper is an effective attempt to popularize complex neural networks and seek new methods of designing more complex FIR filters.