自适应有源噪声控制算法中的次级通道问题研究
文献类型：学位论文
| 作者 | 赵剑 |
| 学位类别 | 博士 |
| 答辩日期 | 2007-06-08 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 声学研究所 |
| 关键词 | 有源噪声控制 自适应滤波 次级通道 不确定性 H2/Hinf 多模型 |
| 其他题名 | Research on the Secondary Path Problem in Adaptive Active Noise Control Algorithms |
| 学位专业 | 信号与信息处理 |
| 中文摘要 | 伴随着经济、社会的发展，低频噪声污染成为日益严重的社会问题。传统的噪声控制技术采用吸声、隔声材料与器件等无源噪声控制方法，一般而言对低频噪声的控制效率不高。有源噪声控制（ANC）技术可以对低频噪声进行有效控制，显示出广阔的应用前景，长期以来一直是非常热门的研究方向。 ANC系统通常包括电声部分和控制器部分，其中控制器一直是ANC系统中的一项关键技术。基于数字电路的自适应ANC系统在控制时变噪声和复杂系统方面相对于由固定系数滤波器或模拟电路实现的控制系统来说体现出巨大的优越性，但成功实现一个自适应ANC系统还是存在很多尚未解决的技术问题，其中最关键的问题之一是当次级通道存在很大不确定性的时候如何保证自适应算法的稳定性。自适应算法需要得到次级通道的相对准确模型（从次级声源到误差传声器之间的传递函数，通常可以通过离线或在线系统辨识方法进行估计），用以更新自适应滤波器的权值。然而，实际系统往往是时变的，次级通道可能存在很大不确定性，这就导致了次级通道及其辨识模型之间存在较大的辨识误差，系统可能存在不稳定性问题。 本论文的研究目标为：针对应用于自适应ANC算法中普遍存在的由于次级通道不确定性所引起的系统可能出现的不稳定性问题，研究了三种可行的次级通道问题解决方案，分别是针对有源抗噪声护耳器系统的基于单个次级通道模型优化的ANC算法，基于多次级通道模型自适应的ANC算法，无次级通道模型ANC算法。分别阐述如下： (1) 针对有源抗噪声护耳器系统处在不同的佩戴松紧程度，导致真实次级通道存在很大不确定性的问题，提出了基于单个次级通道模型优化的ANC算法，采用固定系数和自适应滤波器相结合来实现一种复合式反馈有源护耳器。本方法的创新之处为：固定系数滤波器的设计目标主要是减小护耳器系统真实次级通道的不确定性，从而保证自适应算法的稳定性，而传统方法中其固定系数滤波器设计目标主要是为了降低初级宽带噪声。固定系数滤波器和自适应算法中的次级通道模型都采用 优化方法进行优化设计。理论分析和实验结果表明：固定系数滤波器有效地减小了真实次级通道的不确定性，优化得到的最优次级通道滤波器保证了护耳器佩戴状态发生改变时自适应算法的稳定性。 (2) 引入控制学中的多模型自适应控制的思想，提出了基于多次级通道模型自适应的ANC算法。该算法对真实次级通道建立处在不同状态的多个模型覆盖其不确定性，在系统运行过程中根据预设的切换准则实时判断当前真实次级通道状态并在次级通道模型间进行快速准确切换。针对不同的ANC系统建立不同的切换准则，提出了三种多模型ANC算法的实现方式，分别是仿结构间接多模型ANC、基于Modified FXLMS算法的间接多模型ANC和直接多模型ANC。仿真和实验结果表明：基于多次级通道模型自适应的ANC算法能跟踪真实次级通道的变化并在次级通道模型间进行快速准确的切换，从而保证了自适应算法的稳定性。 (3)无次级通道模型ANC算法采用附加瞬时扰动的方法得到目标函数的梯度, 从而不需要次级通道模型就可以进行控制滤波器的权值更新。讨论了该算法的改进方法并应用于降低初级多频噪声，对算法进行了仿真实验验证。 |
| 英文摘要 | Low frequencies noise pollution becomes a more and more serious social problem together with the economic and social development. Conventional methods of suppressing noise using passive noise control technology such as sound absorptive and isolating materials and devices, generally do not work effectively in the low frequency range. Active Noise Control (ANC) technology, on the contrary, shows great advantages on the low frequencies noise control problem, and has been taken as an active research area with promising applications in noise control engineering for decades. Besides those electroacoustic components and setups, the controller has always been taken as a key issue in an ANC system. The ANC system based on adaptive digital controller shows a greater potential to control the time-variant noise and complex system than its fixed and usually analog counterpart. However there are many technical problems have not yet been fully explored on the successful implementation of an adaptive ANC system, among which one key technical problem is the stability issue of the adaptive algorithms when the real secondary path has large uncertainties. As well known, most of the traditional adaptive ANC algorithms require the relative accurate estimation of the real secondary path, viz. the transfer function from the secondary source to the error sensor which can be estimated using off-line or on-line modeling techniques, to ensure stable adaptive updating of the control filter. In practical applications, the ANC systems may be time-variant, and the real secondary path may have large uncertainties. Therefore, the estimation error between the real and the estimated secondary paths is inevitable, which would severely affect the stability of the noise control system. The focus of the dissertation is to solve the stability problem of the adaptive ANC system caused by the secondary path uncertainties. Three methods are proposed and discussed in depths in this research work, including the ANC algorithm based on single secondary path model optimization and its application to ANR headset, the ANC algorithm based on multiple secondary path models adaptation, and the ANC algorithm without the secondary path model. Those contributions are summarized as follows. (1) The ANC algorithm based on single secondary path optimization is presented to deal with the large uncertain problem of the real secondary path when the ANR headset system is at different wearing positions. The hybrid feedback ANR headset system is implemented with both fixed and adaptive filters. The contribution of this method is the main design objective of the fixed filter is to reduce the uncertainties of the real secondary path and thus guarantee the stability of the adaptive filter, while in the traditional method the main design objective is to suppress the broadband nosie. Both the fixed filter and the filter to model the real secondary path are optimized with the method. Theoretical and experimental results show that the fixed filter effectively reduces the uncertainties of the real secondary path and the optimized secondary path filter guarantees the robustess of the adaptive algorithm under large variations of the ANR headset wearing positions. (2) The ANC algorithm based on multiple secondary path models is proposed by introducing the idea of multiple models adaptive control in the modern control literature. This algorithm sets up multiple models to cover the whole range of the uncertainties of the real secondary path. During the operating process, the algorithm determines the specific situation of the real secondary path and switches those secondary path models rapidly and exactly according to a switching rule. Three approaches of the multiple models ANC algorithm are presented with different switching rules which is designed for different ANC systems, including the indirect multiple models ANC with the similar structure in the modern control literature, the indirect multiple models ANC based on Modified FXLMS algorithm, and the direct multiple models ANC. Computer simulations and experimental results show that the proposed method can track the real secondary path’s variations and switch its models rapidly and exactly, therefore stabilize the ANC algorithm. (3) The algorithm without the secondary path model estimates the gradient of the cost function by injecting a simultaneous perturbation into the system, and then updating the adaptive filter without the secondary path model. The modification of the algorithm and its application to the multiple tone noise control problem are discussed and confirmed by simulations. |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 135 |
| 源URL | [http://159.226.59.140/handle/311008/222]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 赵剑. 自适应有源噪声控制算法中的次级通道问题研究[D]. 声学研究所. 中国科学院声学研究所. 2007. |
入库方式： OAI收割
来源：声学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。