In memory polynomial predistorter design, the coefficient estimation algorithm based on normalized least mean square is sensitive to initialization parameters. A predistorter based on generalized normalized gradient d...In memory polynomial predistorter design, the coefficient estimation algorithm based on normalized least mean square is sensitive to initialization parameters. A predistorter based on generalized normalized gradient descent algorithm is proposed. The merit of the GNGD algorithm is that its learning rate provides compensation for the independent assumptions in the derivation of NLMS, thus its stability is improved. Computer simulation shows that the proposed predistorter is very robust. It can overcome the sensitivity of initialization parameters and get a better linearization performance.展开更多
提出一种基于遗传算法和低阶广义记忆多项式实值神经网络的射频功率放大器数字预失真方法。该方法将遗传算法优化的低阶广义记忆多项式模型与神经网络模型进行级联来增强校正模型与功放失真的匹配程度。它不仅可以提升模型的校正能力,...提出一种基于遗传算法和低阶广义记忆多项式实值神经网络的射频功率放大器数字预失真方法。该方法将遗传算法优化的低阶广义记忆多项式模型与神经网络模型进行级联来增强校正模型与功放失真的匹配程度。它不仅可以提升模型的校正能力,同时可以加快网络的收敛速度。采用60MHz的三载波LTE信号进行实验,通过与实值延时线神经网络模型对比,在收敛速度上有显著提升,同时在邻道功率泄露ACLR指标上有6 d B左右改善。展开更多
基金supported by the National High Technology Research and Development Program of China(2006AA01Z270).
文摘In memory polynomial predistorter design, the coefficient estimation algorithm based on normalized least mean square is sensitive to initialization parameters. A predistorter based on generalized normalized gradient descent algorithm is proposed. The merit of the GNGD algorithm is that its learning rate provides compensation for the independent assumptions in the derivation of NLMS, thus its stability is improved. Computer simulation shows that the proposed predistorter is very robust. It can overcome the sensitivity of initialization parameters and get a better linearization performance.
文摘提出一种基于遗传算法和低阶广义记忆多项式实值神经网络的射频功率放大器数字预失真方法。该方法将遗传算法优化的低阶广义记忆多项式模型与神经网络模型进行级联来增强校正模型与功放失真的匹配程度。它不仅可以提升模型的校正能力,同时可以加快网络的收敛速度。采用60MHz的三载波LTE信号进行实验,通过与实值延时线神经网络模型对比,在收敛速度上有显著提升,同时在邻道功率泄露ACLR指标上有6 d B左右改善。
