电子测试
電子測試
전자측시
ELECTRONIC TEST
2009年
7期
5-8,65
,共5页
复激活函数神经网络%可规划相位%FIR滤波器%误差加权函数值
複激活函數神經網絡%可規劃相位%FIR濾波器%誤差加權函數值
복격활함수신경망락%가규화상위%FIR려파기%오차가권함수치
complex activation function neural networks%planned phase%FIR filters%weighted error function
本文针对可规划相频响应的实系数FIR滤波器的逼近问题,采用一个三层复激活函数前馈神经网络来实现.该网络隐层各神经元的激活函数为复指数函数,将滤波器系数作为隐层各神经元到输出层的连接权值,通过对误差函数的最小化来调整权值,并根据网络特性与所要设计的滤波器的特点,提出了一些实际设计中训练样本集选取与误差加权值设置的规则.依托所采用的神经网络,根据上述规则,进行了两例可规划相频特性的实系数FIR滤波器的设计,结果表明所设计滤波器的相频响应较好地满足了设计要求.
本文針對可規劃相頻響應的實繫數FIR濾波器的逼近問題,採用一箇三層複激活函數前饋神經網絡來實現.該網絡隱層各神經元的激活函數為複指數函數,將濾波器繫數作為隱層各神經元到輸齣層的連接權值,通過對誤差函數的最小化來調整權值,併根據網絡特性與所要設計的濾波器的特點,提齣瞭一些實際設計中訓練樣本集選取與誤差加權值設置的規則.依託所採用的神經網絡,根據上述規則,進行瞭兩例可規劃相頻特性的實繫數FIR濾波器的設計,結果錶明所設計濾波器的相頻響應較好地滿足瞭設計要求.
본문침대가규화상빈향응적실계수FIR려파기적핍근문제,채용일개삼층복격활함수전궤신경망락래실현.해망락은층각신경원적격활함수위복지수함수,장려파기계수작위은층각신경원도수출층적련접권치,통과대오차함수적최소화래조정권치,병근거망락특성여소요설계적려파기적특점,제출료일사실제설계중훈련양본집선취여오차가권치설치적규칙.의탁소채용적신경망락,근거상술규칙,진행료량례가규화상빈특성적실계수FIR려파기적설계,결과표명소설계려파기적상빈향응교호지만족료설계요구.
To the approximation problem of finite impulse response(FIR)filters with planned phase and real coefficients,a three-layer feedforward neural network with complex activation function is used to implement it in this paper.The complex exponential function is used as activation function of neurons in hidden layer,and the coefficients of the actual FIR filter are used as the connection weights between neurons in hidden layer and output layer in the network.It adjusts weights by minimizing the error function.Some practical design rules of which training sets selection and error weight function values setting are proposed based on the characteristics of the network and the desired filters,besides two FIR filters with planned phase are designed based on the network.The result of design shows that the phase response can satisfy the design requirement.
