本文選題：狀態(tài)轉(zhuǎn)移算法　切入點：正弦基函數(shù)　出處：《計算機仿真》2017年02期 　論文類型：期刊論文

【摘要】：為了改善常用濾波器設(shè)計方法存在的不足,用一組正弦基函數(shù)神經(jīng)網(wǎng)絡(luò)線性組合逼近理想濾波器的振幅響應(yīng),用狀態(tài)轉(zhuǎn)移算法優(yōu)化基函數(shù)神經(jīng)網(wǎng)絡(luò)權(quán)值,使得實際濾波器的幅度響應(yīng)逼近理想濾波器的幅度響應(yīng),建立了基于狀態(tài)轉(zhuǎn)移算法的正弦基函數(shù)神經(jīng)網(wǎng)絡(luò)濾波器設(shè)計模型。與窗函數(shù)法設(shè)計的濾波器對比,上述模型具有更好的性能,實驗結(jié)果表明,狀態(tài)轉(zhuǎn)移算法優(yōu)化權(quán)值的方法,克服了傳統(tǒng)基函數(shù)神經(jīng)網(wǎng)絡(luò)模型權(quán)值不易確定,收斂速度慢等缺點。
[Abstract]:In order to improve the shortcomings of the common filter design methods, a group of sinusoidal basis function neural networks are used to approximate the amplitude response of the ideal filter, and the state transfer algorithm is used to optimize the weight of the basis function neural network. The amplitude response of the actual filter approximates the amplitude response of the ideal filter. A sinusoidal basis function neural network filter design model based on the state transfer algorithm is established, which is compared with the filter designed by the window function method. The experimental results show that the state transfer algorithm can optimize the weights of the model, which overcomes the disadvantages of the traditional basis function neural network model, such as the difficulty to determine the weights and the slow convergence speed.
【作者單位】： 新疆大學(xué)網(wǎng)絡(luò)與信息技術(shù)中心;
【分類號】：TN713

【參考文獻】

相關(guān)期刊論文 前5條

1 鄒娟;賈世杰;曾潔;;基于遺傳算法的FIR濾波器設(shè)計[J];大連交通大學(xué)學(xué)報;2010年04期

2 張永華;歐陽煒霞;劉蕾;郭興龍;賴宗聲;王碩;;MEMS集成濾波器技術(shù)[J];傳感器與微系統(tǒng);2008年07期

3 李輝,張安,趙敏,徐琦;粒子群優(yōu)化算法在FIR數(shù)字濾波器設(shè)計中的應(yīng)用[J];電子學(xué)報;2005年07期

4 曾U喺，

本文編號：1632021