QQ客服 官方微博 反馈留言

题名	子带自适应滤波算法在声学回声抵消中的应用研究
作者	路阳
学位类别	博士
答辩日期	2005
授予单位	中国科学院声学研究所
授予地点	中国科学院声学研究所
关键词	声学回声抵消 子带自适应滤波 房间混响时间 语音统计频谱
中文摘要	自适应声学回声抵消（Adaptive AEC）是现代电话会议系统、车载电话等免持通话设备中的关键技术之一，可以为免持通话提供全双工性能。在电话会议系统应用中，房间冲激响应时间过长，造成自适应算法计算量过大和自适应过程收敛缓慢，子带和变换域自适应滤波算法是解决这类问题的主要手段。子带算法在声学回声抵消中主要有三方面的优势，降低计算量，加快对有色信号激励的收敛速度，以及在设计上具有很大的灵活性。但子带算法的固有延时也限制了其在AEC中的应用范围。本文首先回顾了声学回声抵消技术的常用算法以及当前研究的趋势与热点，然后对子带自适应滤波器的核心问题-滤波器组的原理及设计进行了深入的分析，在综合考虑系统延时、计算量以及混叠等因素后，优化设计并实现了满足多方面要求的低延时滤波器组。实现了高效能的广义离散傅立叶变换（GDFT）调制过采样子带自适应滤波器，采用仿真和实际样本与全带算法进行性能对比分析，验证了子带自适应滤波器的在计算量上的巨大优势和对有色噪声激励的具有较快的收敛速度。房间声学中有混响时间随频率上升而减小的特点，以及语音信号长时平均频谱在高频带逐渐衰减的特点。本文利用子带算法的灵活性，对传统方法加以改进，为不同子带的自适应滤波器分配不同的阶长，在基本保持原有回声抵消性能的基础上，进一步降低了计算量与所需系统资源，为声学回声抵消算法的实时实现提供了有力的支撑。
英文摘要	Adaptive acoustic echo cancellation (AEC) supplies the full-duplex feature in hands-free telecommunication, such as teleconference, hands-free car-kit, etc. In application of teleconference, the long time reverberation time, e.g., the long time impulse response of the large room, results in big problems of adaptive algorithm, such as the huge computation complexity and the slow convergence speed. The subband adaptive filtering (SAF) and transform domain adaptive filtering have been widely used to solve such problems in this field. SAF has three mojor advantages in AEC, which are the reduction of the computation complexity, improvement of the convergence rate and additional degree of freedom offered to the designer. However, the inherent delay of the SAF algorithm limits its application in AEC. After the review of the conventional algorithms of AEC, the optimization processes of the filterbanks and the prototype filter have been discussed in detail in this thesis. With consideration of the overall performance of the SAF algorithm, the compromise has been reached among the computation, the aliasing and the delay to optimize the filterbanks. The oversampled General Discrete Fourier Transform (GDFT) Oversampled SAF has been implemented, and then the performance comparison between the SAF and the fuUband adaptive filter has been carried out. The advantages and disadvantages of the SAF are discussed thoroughly thereafter. Finally, a modified algorithm is presented, which makes use of the acoustic characteristics of rooms and speech. Statistically, the reverberation durations are shorter in middle and high frequency range, and so are the long-term average speech spectra. Based on above acoustic evidences, the SAF algorithm has been modified to allocate different adaptive filter taps in different subband to achieve much less computation complexity and resources, while keeping the same performance as the traditional SAF.
语种	中文
公开日期	2011-05-07
页码	103
内容类型	学位论文
源URL	[http://159.226.59.140/handle/311008/964]
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	路阳. 子带自适应滤波算法在声学回声抵消中的应用研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2005.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

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。