资源简介
关于稀疏分解的盲源分离程序,有相应的文章可相对照

代码片段和文件信息
function [y A]= sparseBSS1(XLlangdaGhdelta)
%----------------------------------------------------------------
% 2009-04-15 YangZhicong
% X: observed signaleach row correspond to a sensor observations
% L: the length of FFT (or the length of Hanning window)
% langda: adjust the desired angular width
% G: % discretize the potential field by taking a sample of G points
% h: threshold
% y: estimated sourses
% A: estimated mix matrix
% here we just consider the special casei.e m = 2
[m T] = size(X);
if nargin < 2L = 2048; end
d = round(0.15*L)*2; % the hop distance
overlap = L - d; % number of samples of overlap between adjacent windows
w = hann(L)‘; % Hann (Hanning) window function;
frame_X = bss_make_frames(Xwoverlap); % decompose X into frames. For instance frame_X(::1)
% correspond to 1st sensoreach row is a frame.
% now each column of frame_X(::i) is a frame for convenience of fft operation
for i=1:m
frame_X_temp(::i) = frame_X(::i)‘;
end
frame_X = frame_X_temp;
clear frame_X_temp;
frame_X_Fs = fft(frame_X);
K = L/2 +1; % preserve the positive half spectrum
for i = 1:m
temp_Matrix = frame_X_Fs(1:K:i);
realPart = real(temp_Matrix);
imagPart = imag(temp_Matrix);
temp_Matrix = [realPart;imagPart];
[rowcolumn] = size(temp_Matrix);
Xu(i:) = reshape(temp_Matrix 1row*column);
end
if nargin < 4 G =60; end
if nargin < 3langda = 5; end
% we consider the special case for m = 2
theta = rem(atan2(Xu(2:)Xu(1:))+2*pipi);
radius = sqrt((Xu(1:).*Xu(1:)+Xu(2:).*Xu(2:)));
radius = radius /max(radius); % normalize to one
theta_k = pi/2/G + (1:G)*pi/G;
if nargin < 5h = 0.2;end % discarding the less reliable data points
index = find(radius>=h);
radius = radius(index);
theta = theta(index);
figure(‘Name‘‘Scatter Plot ‘);
polar(thetaradius‘.‘);
for k = 1:G
potential(k) = sum(radius.*Tao(langda*(theta-theta_k(k))));
end
figure(‘Name‘‘Potential Funtction‘);
plot(theta_k/pi*180potential);
% A point is considered a maximum peak if it has the maximal
% value and was preceded (to the left) by a value lower by
% DELTA.
if nargin < 6 delta = max(abs(potential))*0.1; end
maxtab = peakdet(potential delta theta_k); % find the maxima of potential function
theta_esti = maxtab(:1);
% get the first m maxima of potential function
theta_esti = sort(theta_esti‘descend‘);
theta_esti = theta_esti (1:m);
A = [cos(theta_esti) sin(theta_esti)]‘;
W = inv(A); % separation matrix
y = W*X;
end
%%
% ------------------------------PEAKDET.m--------------------------------
function [maxtab mintab]=peakdet(v delta x)
%PEAKDET Detect peaks in a vector
% [MAXTAB MINTAB] = PEAKDET(V DELTA) finds the local
% maxima and minima (“peaks“) in the vector V.
% MAXTAB and MINTAB consists of two columns. Column 1
% contains indices in V and column 2 the
属性 大小 日期 时间 名称
----------- --------- ---------- ----- ----
文件 4590 2009-04-16 12:31 稀疏盲源分离自编程序\sparseBSS1.m
文件 19968 2011-10-11 19:59 稀疏盲源分离自编程序\sparseBSS1说明.doc
目录 0 2011-10-11 20:02 稀疏盲源分离自编程序
----------- --------- ---------- ----- ----
24558 3
相关资源
- 稀疏表示SRC算法
- SfM稀疏三维点云重建--完整工程文件
- 空间稀疏预编码很好的文章
- 阵列测向的稀疏超分辨方法研究
- 基于稀疏分解的图像修复MP
- 详解稀疏编码百度深度学习--ppt
- 基于稀疏表示的人脸识别文章及对应
- CUDA实现稀疏大矩阵乘法
- 基于稀疏A*航迹规划算法
- 斯坦福大学基于稀疏的统计学习
- RPCA 低秩稀疏分解
- 稀疏和冗余-在信号和图像处理方面从
- 稀疏表示在图像处理中的应用
- 多尺度稀疏 图像融合.zip
- 基于稀疏表示的图像超分辨PPT
- 压缩感知基础
- 一种有效的稀疏编码源码
- 稀疏信号与压缩感知系列讲座PPT李廉
- 基于稀疏表示的图像去噪算法研究
- Statistic learning with sparsity
- Sparse Subspace Clustering: Algorithm Theory a
- 频域卷积混合盲源分离
- Sparse and Redundant Representations 中文翻译
- 稀疏表示和协同编码哪个对人脸识别
- 基于稀疏表示的图像去噪算法
- 稀疏冗余表示 在信号和图像处理方面
- 基于稀疏表示的图像超分辨方法研究
- KSVD稀疏表示字典训练程序
- 稀疏和冗余-在信号和图像处理方面从
- 稀疏分解图像去噪
评论
共有 条评论