moving.m

我认为很不错的语音处理的ｍａｔｌａｂ源代码

function [y]=moving(x,m,fun)
%MOVING will compute moving averages of order n (best taken as odd)
%
%Usage: y=moving(x,n[,fun])
%where x 	is the input vector (or matrix) to be smoothed. 
%      m 	is number of points to average over (best odd, but even works)
%      y 	is output vector of same length as x
%      fun  (optional) is a custom function rather than moving averages
%
% Note:if x is a matrix then the smoothing will be done 'vertically'.
% 
%
% Example:
%
% x=randn(300,1);
% plot(x,'g.'); 
% hold on;
% plot(moving(x,7),'k'); 
% plot(moving(x,7,'median'),'r');
% plot(moving(x,7,@(x)max(x)),'b'); 
% legend('x','7pt moving mean','7pt moving median','7pt moving max','location','best')
%
% optimized Aslak Grinsted jan2004
% enhanced Aslak Grinsted Apr2007


if m==1
    y=x;
    return
end
if size(x,1)==1
    x=x';
end

if nargin<3
    fun=[];
elseif ischar(fun)
    fun=eval(['@(x)' fun '(x)']);
end

if isempty(fun)

    f=zeros(m,1)+1/m;
    n=size(x,1);
    isodd=bitand(m,1);
    m2=floor(m/2);


    if (size(x,2)==1)
        y=filter(f,1,x);
        y=y([zeros(1,m2-1+isodd)+m,m:n,zeros(1,m2)+n]);
    else
        y=filter2(f,x);
        y(1:(m2-~isodd),:)=y(m2+isodd+zeros(m2-~isodd,1),:);
        y((n-m2+1):end,:)=y(n-m2+zeros(m2,1),:);
    end

else
    y=zeros(size(x));
    sx=size(x,2);
    x=[nan(floor(m*.5),sx);x;nan(floor(m*.5),sx)];
    m1=m-1;
    for ii=1:size(y,1);
        y(ii,:)=fun(x(ii+(0:m1),:));
    end
    
end

return