自适应滤波的应用课件.ppt
自适应滤波的应用课件 Still waters run deep.流静水深流静水深,人静心深人静心深 Where there is life, there is hope。有生命必有希望。有生命必有希望多麦克风降噪参考麦克风输入主麦克风输入噪声源MATLAB 处理同学们好模型S 信号n0 噪声,它是n1经过滤波以后得到的噪声,即n0 =filter(w, n1)n1 参考噪声y n1 经过滤波以后得到的信号(滤波器w的系数未知)假设假设: 它们都是零均值的, S与n0及n1不相关, 但是n0 和n1相关Singal + Noise0FilterNoise1s+n0n1yRecovered Signalz+-主麦克风输入参考麦克风输入模型目标: 对n1进行自适应滤波,滤波器为w, 从主麦克风得到的信号中减去y, 得到 “干净”的信号z,从而达到降噪 的目的。Singal + Noise0FilterNoise1s+n0n1yRecovered Signalz+-主麦克风输入参考麦克风输入算法)(20nyE 20min22minnyEsEzE代价函数策略策略: 调整滤波器系数, 使 最小, 这也意味着 最小. 这个意义下, 可以看成是误差 信号z2zE 2202022nyEsEnysEsEzE滤波器系数的调整滤波器的阶数是worder, 滤波器的输出其中调整滤波器系数的LMS算法Tinworderini)(,),1()(111n)()()(1iiiyTnw)()() 1()(1iziiinww滤波器系数的调整调整滤波器系数的RLS算法)()() 1()(iziiikww)() 1()(1)() 1()(11111iiiiiiTnPnnPk) 1()()() 1()(111iiiiiTPnkPP语音00.511.522.533.544.55x 104-0.6-0.4-0.200.20.40.6signal噪声环境中的语音00.511.522.533.544.55x 104-10-50510primary microphone signalLMS滤波效果00.511.522.533.544.55x 104-4-2024primary microphone signal00.511.522.533.544.55x 104-101filtered output0123456x 1040102030Mean Squared Error主麦克风信号滤波输出信号均方误差RLS滤波效果00.511.522.533.544.55x 104-505primary microphone signal00.511.522.533.544.55x 104-0.500.5filtered output0123456x 104024Mean Squared Error主麦克风信号滤波输出信号均方误差线性调频信号0200400600800100012001400160018002000-1-0.8-0.6-0.4-0.200.20.40.60.81signalLMS滤波效果主麦克风信号滤波输出信号均方误差0200400600800100012001400160018002000-5050200400600800100012001400160018002000-1010500100015002000250001020RLS滤波效果主麦克风信号滤波输出信号均方误差0200400600800100012001400160018002000-5050200400600800100012001400160018002000-101050010001500200025000510
