基于支持向量机的记忆功率放大器预失真

提出了一种新的针对记忆非线性功率放大器的支持向量机（SVM）预失真器。通过对其建模中采用径向基核函数和多项式核函数所表现出的性能特点进行分析,为核函数的选取提供了参考。采用以多项式为核函数的SVM对3种典型的记忆非线性功率放大器模型进行线性化仿真,结果表明了该方法的有效性和鲁棒性。     

记忆非线性功率放大器的高效预失真

记忆非线性放大器的预失真问题一直是预失真技术的难点。通常采用Volterra级数、Hammerstein模型和神经网络等模型的记忆预失真都存在形式复杂、自适应困难的缺点。文章通过增加两个延时环节将基于多项式的无记忆放大器的高效预失真结构推广到有记忆放大器的预失真中,并联合一种简单的带抽头延时的非线性多项式模型作为记忆预失真器模型实现了记忆非线性放大器的快速、高效的线性化。仿真结果表明,利用所提出的预失真方案能快速实现记忆放大器的预失真,而且显著提高了线性化性能。     

基于直接学习法的一种新预失真器

为了克服宽带信号经过记忆放大器的非线性失真,针对有记忆非线性功放的多项式模型,提出了一种新的基于直接学习法的自适应算法.该算法采用无记忆预失真器的级联扩展,具有横向滤波器结构,与记忆多项式有相似的线性化效果.并且针对信号噪声对自适应算法的扰动和收敛速度慢等缺点,采用归一化LMS算法加以改进.在非线性功放的记忆多项式模型下,通过宽带信号验证了基于直接学习法的记忆型预失真器算法的有效性.     

A novel distortion elimination method of power amplifier in wideband OFDM system

The nonlinear,memory and saturate distortion caused by power amplifier lead to low BER per-formance and low power efficiency in wideband OFDM system.To solve the related problems,a novel power amplifier based on combined amplification with multi power amplifiers is proposed in this paper.In addition,a novel simplified filter look-up table predistorter is proposed.The novel predistorter connects to a memoryless predistorter subsystem with an adaptive filter subsystem in cascade,and can eliminate the nonlinear memory distortion effectively.The simulation results indicate that the power efficiency and BER performance will be improved by using the novel amplifier structure together with the novel predistorter.     

一种宽带OFDM系统放大器的失真消除方法

放大器非线性、记忆性和饱和性失真导致宽带OFDM系统误码率性能和功率效率低.为解决上述问题,文中提出基于多个小动态范围放大器联合放大的放大器动态范围扩展方法,纠正放大器的饱和性失真;同时,提出一种简化滤波查表预失真方法,该方法由一个无记忆预失真器子系统串联一个自适应滤波器子系统构成,能有效的纠正放大器有记忆非线性失真.仿真表明,两种方法的联合使用可以有效的纠正放大器失真带来的影响,提高系统功率效率和误码率性能.     

基于分片线性化方法的非线性系统多模型自适应控制

基于分片线性化方法辨识一类非线性系统 ,给出了非线性系统的多线性模型表示。基于线性模型建立多个控制器 ,基于最大最小指标切换函数构成多模型自适应控制器。给出了非线性系统多模型自适应控制算法的优化模型集建立方法 ,解决了多模型自适应控制模型多、计算量大的问题。仿真结果证明了算法的有效性     

基于并行演化计算的记忆非线性功率放大器数字预失真研究

自适应数字预失真是克服高功率放大器非线性失真最有前途的一项技术。为提高预失真的效率和效果,引入并行计算平台下的演化计算技术,提出了基于PSO算法预训练神经网络的方法,给出了算法软件实现的基本流程。在所述基础上,采用带抽头延时的双入双出三层前向神经网络结构,根据非直接学习结构和反向传播算法实现自适应,可同时补偿放大器的记忆失真和非线性失真的预失真技术。仿真实验表明,通过与无PSO预训练算法的相比,基于PSO预训练的神经网络训练算法有更好的性能。     

An open‐loop digital predistorter based on memory polynomial inverses for linearization of RF power amplifier

Digital Predistorter is a cost‐effective solution to compensate for the nonlinear distortions appearing in the RF power amplifiers (PAs). The indirect learning scheme is widely implemented because of its flexibility to eliminate the requirement for building a closed‐loop real time system, which dramatically reduces the complexity for measurement setup. However, such scheme is sensitive to the measurement noise that may cause biasing in the coefficients estimation. To minimize the influence of measurement noise and simultaneously enable the open‐loop implementation, we propose a predistortion technique that first model the PA and then generates predistorted signal iteratively through a feedback configured structure to avoid using the noisy signal when performing the inverse model estimation. Unlike the indirect learning which estimates the postinverse of the PA, our predistortion is based on the preinverse of the PA. Both simulations and measurements show that utilizing the proposed predistortion can obtain adjacent channel power ratio (ACPR) improvement in wideband code division multiple access (WCDMA) signal test compared with the conventional memory polynomial predistortion based on indirect learning. © 2011 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2011.     

P阶逆算法在非线性信道中的应用

针对数字无线通信中高功率放大器的非线性失真,提出了一种Volterra滤波均衡器的设计方法一阶逆算法,即对原有的Volterra级数模型进行求逆的过程。而根据该均衡器所处的位置又可以分为阶逆前向均衡和阶逆后向均衡,并对阶逆前向均衡的结构和算法进行了研究。计算机模拟结果表明,采用该算法系统输出误码率大为降低,整个系统性能具有明显的改善,且计算量小,可实现性强。     

Modeling of wideband radio frequency power amplifiers using Zernike polynomials

In this article, we suggest a new set of basis functions that are based on Zernike polynomials for the behavioral modeling of radio frequency power amplifiers (PAs). The modeling of highly nonlinear PAs exhibits numerical instability that degrades the accuracy of the model parameters and predistorter modeling efficiency. Simulation results show that the proposed polynomial model is more suitable to resolve the numerical instability problem and proves to have greater accuracy with reduced complexity. A Doherty PA driven by a multicarrier wideband code division multiple access signal was used for validation; and, the obtained results show that the new model exhibits superior numerical stability as the nonlinearity order and memory depth of the model increase. © 2012 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2012.     

预失真线性化器的仿真研究

根据级联网络理论,基于行波管的非线性特性推导出预失真器模型,并在Matlab仿真平台上从三个角度验证了该模型的可行性和有效性,包括增益特性、相移特性和三阶交调特性,其中三阶交调系数的提取采用了计算精度高的自相关函数法。仿真结果显示,在饱和输入功率条件下加入预失真器后的系统相比行波管的三阶交调改善了6．947dBc。     

并行单型规范分片线性函数极坐标数字预失真器

介绍了一种并行的基于单型规范分片线性（SCPWL）函数的极坐标数字预失真器模型。首先根据分片线性拟合原理引入绝对值型的单型规范分片线性函数,建立模型解析方程并推导出模型参数的最小二乘解,然后根据模型特点设计了数字预失真器的并行实现架构,并对模型架构的工作顺序和参数提取过程进行了介绍。仿真结果表明SCPWL极坐标预失真器远远优于传统的功率回退线性化技术,且与常用的复多项式预失真器相比较,尤其是在过饱和非线性失真情况下,其线性补偿能力要优于后者。     

一类非线性动态系统基于强化学习的最优控制制

提出一类非线性不确定动态系统基于强化学习的最优控制方法. 该方法利用欧拉强化学习算法估计对象的未知非线性函数, 给出了强化学习中回报函数和策略函数迭代的在线学习规则. 通过采用向前欧拉差分迭代公式对学习过程中的时序误差进行离散化, 实现了对值函数的估计和控制策略的改进. 基于值函数的梯度值和时序误差指标值, 给出了该算法的步骤和误差估计定理. 小车爬山问题的仿真结果表明了所提出方法的有效性.

     

宽带GaN功率放大器的建模与预失真

对正交频分复用传输系统中宽带GaN功率放大器的建模与预失真线性化方法进行研究。应用一个带有记忆的多项式模型来建模GaN功率放大器及其逆特性,并通过递推最小二乘法(RLS)辨识模型参数,然后利用预失真间接结构实现了GaN功率放大器的预失真器。最后,仿真验证预失真方法的有效性,结果表明记忆多项式模型可以对GaN功率放大器进行建模并实现线性化。     

Performance analysis of ASR system in hybrid DNN-HMM framework using a PWL euclidean activation function

Automatic Speech Recognition (ASR) is the process of mapping an acoustic speech signal into a human readable text format. Traditional systems exploit the Acoustic Component of ASR using the Gaussian Mixture Model- Hidden Markov Model (GMM-HMM) approach.Deep NeuralNetwork (DNN) opens up new possibilities to overcome the shortcomings of conventional statistical algorithms. Recent studies modeled the acoustic component of ASR system using DNN in the so called hybrid DNN-HMM approach. In the context of activation functions used to model the non-linearity in DNN, Rectified Linear Units (ReLU) and maxout units are mostly used in ASR systems. This paper concentrates on the acoustic component of a hybrid DNN-HMM system by proposing an efficient activation function for the DNN network. Inspired by previous works, euclidean norm activation function is proposed to model the non-linearity of the DNN network. Such non-linearity is shown to belong to the family of Piecewise Linear (PWL) functions having distinct features. These functions can capture deep hierarchical features of the pattern. The relevance of the proposal is examined in depth both theoretically and experimentally. The performance of the developed ASR system is evaluated in terms of Phone Error Rate (PER) using TIMIT database. Experimental results achieve a relative increase in performance by using the proposed function over conventional activation functions.     

Harmonic solution of semiconductor transport equations for microwave and millimetre-wave device modelling

The hydrodynamic transport equations for charges in a semiconductor have been solved for a periodic excitation by means of a harmonic approach, in order to model microwave and millimetre-wave active devices. The solution is based on the expansion of physical variables in a Fourier series in the time domain, and on discretisation in the space domain. A waveform-balance technique in the TD is used to solve the nonlinear equations system. This approach allows for a longer time step with respect to standard TD solutions for most cases of interest, greatly reducing simulation time by at least two orders of magnitude in typical cases. © 2003 Wiley Periodicals, Inc. Int J RF and Microwave CAE 14: 36–48, 2004.     

Model predictive control for constrained uncertain piecewise linear systems

For constrained piecewise linear (PWL) systems, the possible existing model uncertainty will bring the difficulties to the design approaches of model predictive control (MPC) based on mixed integer programming (MIP). This paper combines the robust method and hybrid method to design the MPC for PWL systems with structured uncertainty. For the proposed approach, as the system model is known at current time, a free control move is optimized to be the current control input. Meanwhile, the MPC controller uses a sequence of feedback control laws as the future control actions, where each feedback control law in the sequence corresponds to each partitions and the arbitrary switching technique is adopted to tackle all the possible switching. Furthermore, to reduce the online computational burden of MPC, the segmented design procedure is suggested by utilizing the characteristics of the proposed approach. Then, an offline design algorithm is proposed, and the reserved degree of freedom can be online used to optimize the control input with lower computational burden.     

Adaptive Predistortion and Postdistortion for Nonlinear Channel

This paper proposes a new adaptive predistortion-postdistortion scheme based on a recurrent neural network to reduce nonlinear distortion introduced by a high power amplifier in the amplitude and phase of received Quadrature Phase Shift Keying (QPSK) signals in a digital microwave system. The recurrent neural network structure is inspired by the model proposed by Williams and Zipser, with a modified backpropagation algorithm. The input signal is processed by a nonlinear predistorter which reduces the warping effect. The received output signal is passed through a postdistorter to compensate for the warping and clustering effects produced by an amplifier. The proposed scheme yields a significant improvement when it is compared to the system without predistortion-postdistortion, perform-ance is evaluated in terms of the bit error rate and output signal constellation.     

Reinforcement learning in continuous time and space

This article presents a reinforcement learning framework for continuous-time dynamical systems without a priori discretization of time, state, and action. Based on the Hamilton-Jacobi-Bellman (HJB) equation for infinite-horizon, discounted reward problems, we derive algorithms for estimating value functions and improving policies with the use of function approximators. The process of value function estimation is formulated as the minimization of a continuous-time form of the temporal difference (TD) error. Update methods based on backward Euler approximation and exponential eligibility traces are derived, and their correspondences with the conventional residual gradient, TD(0), and TD(lambda) algorithms are shown. For policy improvement, two methods-a continuous actor-critic method and a value-gradient-based greedy policy-are formulated. As a special case of the latter, a nonlinear feedback control law using the value gradient and the model of the input gain is derived. The advantage updating, a model-free algorithm derived previously, is also formulated in the HJB-based framework. The performance of the proposed algorithms is first tested in a nonlinear control task of swinging a pendulum up with limited torque. It is shown in the simulations that (1) the task is accomplished by the continuous actor-critic method in a number of trials several times fewer than by the conventional discrete actor-critic method; (2) among the continuous policy update methods, the value-gradient-based policy with a known or learned dynamic model performs several times better than the actor-critic method; and (3) a value function update using exponential eligibility traces is more efficient and stable than that based on Euler approximation. The algorithms are then tested in a higher-dimensional task: cart-pole swing-up. This task is accomplished in several hundred trials using the value-gradient-based policy with a learned dynamic model.     

Modulo Path History for the Reduction of Pipeline Overheads in Path-based Neural Branch Predictors

Neural-inspired branch predictors achieve very low branch misprediction rates. However, previously proposed implementations have a variety of characteristics that make them challenging to implement in future high-performance processors. In particular, the path-based neural predictor (PBNP) and the piecewise-linear (PWL) predictor require deep pipelining and additional area to support checkpointing for misprediction recovery. The complexity of the PBNP predictor stems from the fact that the path history length, which determines the number of tables and pipeline stages, is equal to the history length, which is typically very long for high accuracy. We propose to decouple the path-history length from the outcome-history length through a new technique called modulo-path history. By allowing a shorter path history, we can implement the PBNP and PWL predictors with significantly fewer tables and pipeline stages while still exploiting a traditional long branch outcome history.