Cancellation of nonlinear distortion based on integration of FCM clustering algorithm and adaptive-t

• 作 者: WANG Gui-ye ZOU Wei-xia WANG Zhen-yu DU Guang-long GAO Ying
• 机 构: Key laboratory of wireless universal communications, Beijing University of Posts and Telecommunicati
• 出 处: The Journal of China Universities of Posts and Telecommunica
• 发表日期: 2014年03期
• 关键词: clustering eliminate cancellation amplifier distortion similarity trained invers

A hybrid system of the fuzzy c-means(FCM)clustering algorithm and adaptive-two-stage linear approximation was presented for nonlinear distortion cancellation of radio frequency(RF)power amplifier(PA).This mechanism can effectively eliminate noise,adaptively model PA’s instantaneous change,and efficiently correct nonlinear distortion.This article puts forward the FCM clustering algorithm for clustering received signals to eliminate white noise,and then uses the adaptive-two-stage linear approximation to fit the inverse function of the amplitude’s and phase’s nonlinear mapping during the training phase.Parameters of the linear function and similarity function are trained using the gradient-descent and minimum mean-square error criteria.The proposed approach’s training results is directly employed to eliminate sampling signal’s nonlinear distortion.This hybrid method is realized easier than the multi-segment linear approximation and could reduce the received signal’s bit error rate(BER)more efficiently.

A hybrid system of the fuzzy c-means (FCM) clustering algorithm and adaptive-two-stage linear approximation was presented for nonlinear distortion cancellation of radio frequency (RF) power amplifier (PA) .This mechanism can effectively eliminate noise, adaptively model PA’s instantaneous change, and efficiently correct nonlinear distortion. This article puts forward the FCM clustering algorithm for clustering received signals to eliminate white noise, and then uses the adaptive-two-stage linear approximation to fit the inverse function of the amplitude’s and phase’s nonlinear mapping during the training phase. Parameters of the linear function and similarity function are trained using the gradient-descent and minimum mean-square error criteria. The proposed approach’s training results is directly employed to eliminate the sampling signal’s nonlinear distortion. This hybrid method is realized easier than the multi-segment linear approximation and could reduce the received signal’s bit error rate (BER) more efficiently.