-
Notifications
You must be signed in to change notification settings - Fork 22
/
FastKalmanFilter.m
69 lines (58 loc) · 2.38 KB
/
FastKalmanFilter.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
function [output,Gp,erle] = FastKalmanFilter(lamda,delta,Np,input,target)
% Ensure row vectors ------------------------------------------------------
s=size(input); if s(1)>s(2), input=input.'; end
s=size(target); if s(1)>s(2), target=target.'; end
% Initialization ----------------------------------------------------------
weight = zeros(Np,1); % filter weight
inputLength = length(input); % length of input signal
output = zeros(1,inputLength); % filter output
inputSequence = zeros(1,Np); %
gain = zeros(Np+1,1); %
weightFor = zeros(Np,1); %
weightBac = zeros(Np,1); %
factor = delta*lamda^-2;
% Filtering ---------------------------------------------------------------
d_sum = 0;
deltad_sum = 0;
%% Do Onling Learning
fprintf('## Do online learning, Please wait... \n');
%>>>>>>>>>>>> Set the waitbar - Initialization <<<<<<<<<<<<<<<<<<
wb1 = waitbar(0, 'FKF Online Training in Progress...');
for i=1:inputLength
%>>>>>>>>>>>>>>>>> Display Waitbar <<<<<<<<<<<<<<<<<<<<<<
waitbar(i/inputLength,wb1)
set(wb1,'name',['Progress = ' sprintf('%2.1f',i/inputLength*100) '%']);
%>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
% prior
fpri = input(i)-inputSequence*weightFor;
weightFor = weightFor+fpri*gain(1:Np);
% post
fpos = input(i)-inputSequence*weightFor;
factor = lamda*factor+fpri'*fpos;
gain = [0;gain(1:Np)]+(fpos'/factor)*[1;-weightFor];
uiN = inputSequence(Np);
inputSequence = [input(i),inputSequence(1:Np-1)];
bpri = uiN-inputSequence*weightBac;
gain(1:Np) = (gain(1:Np)+gain(Np+1)*weightBac)/(1-gain(Np+1)*bpri);
weightBac = weightBac+bpri*gain(1:Np);
% Compute error
output(i) = inputSequence*weight;
preError(i) = target(i)-output(i);
% Update filter weight
weight = weight + gain(1:Np)*preError(i);
%--- Performance measurement
d_sum = d_sum + target(i)^2;
deltad_sum = deltad_sum + preError(i)^2;
erle(i) = 10*log10((d_sum+eps)/(deltad_sum+eps));
end
close(wb1);% close waitbar.
plotEnable = 1;
if plotEnable == 1
figure;
plot(target,'b') ; hold on; grid on;
plot(output,'r');
plot(preError,'g');
title('training: teacher sequence (blue) vs predicted sequence (red)') ;
end
%% Gp
Gp = 10*log10(sum(target.^2)/sum(preError.^2));