forked from ysparkwin/AlternatingProjections
-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy patha1_NUA_SNR.m
322 lines (253 loc) · 11 KB
/
a1_NUA_SNR.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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
% Version 1.0: (05/09/2022)
% written by Y. Park
% System framework as defined in Noiselab DOA estimation
% Version 2.0: (06/08/2022)
% written by Y. Park
% AP Covariance (JASA 2022)
% Mark Wagner, Yongsung Park, & Peter Gerstoft
% MPL/SIO/UCSD
% noiselab.ucsd.edu
% Citation
% M. Wagner, Y. Park, and P. Gerstoft, “Gridless DOA estimation and root-MUSIC for non-uniform linear arrays,” IEEE Trans. Signal Process. 69, 2144–2157 (2021).
% M. Wagner, P. Gerstoft, and Y. Park, “Gridless DOA estimation via alternating projections,” in Proc. IEEE ICASSP (2019), pp. 4215–4219.
% Y. Park and P. Gerstoft, “Alternating pro jections gridless covariance-based estimation for DOA,” in Proc. IEEE ICASSP (2021), pp. 4385–4389.
% Y. Park and P. Gerstoft, “Gridless sparse covariance-based beamforming via alternating projections including co-prime arrays,” J. Acoust. Soc. Am. 151(6), 3828-3837 (2022).
%%
clear; clc; %close all;
addpath([cd,'/_common'])
errCut = 10; % Maximum RMSE cut-off.
dbstop if error;
% Case: NUA Multi- snapshot (for NUA, see Line 47-49)
Nsnapshot = 20;
% anglesTrue = [-65; -2; 3]; % +- .5 Line 73
xAxes = [-10 0 10 20 30 40]; % SNR list
for xaxis = 1:length(xAxes)
% Number of Monte-Carlo simulations
Nsim = 100;
for nsim=1:Nsim
rngN = (xaxis-1)*Nsim + nsim + 100; rng(rngN);
disp(' ')
disp(['SNR',num2str(xAxes(xaxis)),'#Sim : ',num2str(nsim)])
% Environment parameters
c = 1500; % speed of sound
f = 200; % frequency
lambda = c/f; % wavelength
% ULA-horizontal array configuration
Nsensor = 20; % number of sensors
d = 1/2*lambda; % intersensor spacing
q = (0:1:(Nsensor-1))'; % sensor numbering
% sensor spacing perturbation for NUA
qPert = rand(size(q))-.5; qPert(1) = 0; qPert(end) = 0;
q = q+qPert;
xq = q*d; % sensor locations
% signal generation parameters
SNR = xAxes(xaxis);
% total number of snapshots
% Nsnapshot = 10;
% range of angle space
thetalim = [-90 90];
% Angular search grid
theta_separation = .25;
theta = (thetalim(1):theta_separation:thetalim(2))';
Ntheta = length(theta);
% Design/steering matrix (Sensing matrix)
sin_theta = sind(theta);
sensingMatrix = exp(-1i*2*pi/lambda*xq*sin_theta.')/sqrt(Nsensor);
% Generate received signal
anglesTrue = [-65; -2; 3];
anglesTrue = anglesTrue + rand(size(anglesTrue)) - 0.5;
disp(['True DOAs : ',num2str(anglesTrue.')])
% source_amp = [ 7; 7; 7; 4; 4; 13];
anglesTracks = repmat(anglesTrue,[1,Nsnapshot]);
sinAnglesTracks = sind(anglesTracks);
Nsource = numel(anglesTrue);
receivedSignal = zeros(Nsensor,Nsnapshot);
for snapshot = 1:Nsnapshot
% Source amplitude
% source_amp(:,snapshot) = 6*rand(size(anglesTrue)) + 4;
source_amp(:,snapshot) = 10*ones(size(anglesTrue));
% source_amp(:,snapshot) = [10; 7; 4; 7];
Xsource = source_amp(:,snapshot).*exp(1i*2*pi*rand(Nsource,1)); % random phase
% Represenation matrix (steering matrix)
transmitMatrix = exp( -1i*2*pi/lambda*xq*sinAnglesTracks(:,snapshot).' );
% Received signal without noise
receivedSignal(:,snapshot) = sum(transmitMatrix*diag(Xsource),2);
% add noise to the signals
rnl = 10^(-SNR/20)*norm(Xsource);
nwhite = complex(randn(Nsensor,1),randn(Nsensor,1))/sqrt(2*Nsensor);
e = nwhite * rnl; % error vector
receivedSignal(:,snapshot) = receivedSignal(:,snapshot) + e;
% for CRB
crnl(snapshot) = rnl;
cX(:,snapshot) = Xsource;
end
%% CRB
% CRB-YP Van Trees Book Eq.(8.106) & (8.110)
vanTreeV = exp( -1i*2*pi/lambda*xq*sinAnglesTracks(:,snapshot).' );
vanTreeD = (-1i*2*pi/lambda*xq*cosd(anglesTracks(:,snapshot)).') ...
.* exp( -1i*2*pi/lambda*xq*sinAnglesTracks(:,snapshot).' ); % D Eq.(8.100)
Xs = cX;
Pn = mean(diag(((e.*sqrt(crnl/crnl(1)))*(e.*sqrt(crnl/crnl(1)))')/Nsnapshot));
vanTreeSf = diag(diag(Xs*Xs'/Nsnapshot)); % S_f
% H Eq.(8.101) where P_V Eq.(8.96)
H = vanTreeD'...
*(eye(Nsensor) - vanTreeV/(vanTreeV'*vanTreeV)*vanTreeV')...
*vanTreeD;
% det. CRB Eq.(8.110)
CRB = real(H .* (vanTreeSf.'));
CRB = eye(size(Xs,1)) / CRB * (Pn / Nsnapshot / 2);
outputsCRBd(xaxis,nsim) = mean(diag(CRB));
%% Conventional beamforming (CBF)
Ryy = receivedSignal*receivedSignal' / Nsnapshot;
Pcbf = zeros(numel(theta),1);
for ii=1:length(theta)
Pcbf(ii) = sensingMatrix(:,ii)'*Ryy*sensingMatrix(:,ii)/(sensingMatrix(:,ii)'*sensingMatrix(:,ii));
end
% Pcbf = sensingMatrix' * receivedSignal;
% plot(theta,mean(Pcbf.*conj(Pcbf),2)/max(mean(Pcbf.*conj(Pcbf),2)),'k:','linewidth',1,'displayname','CBF')
% plot(theta,abs(Pcbf)/max(abs(Pcbf)),'k:','linewidth',1.5,'displayname','CBF')
[~, Ilocs] = findpeaks(abs(Pcbf),'SORTSTR','descend','Npeaks', Nsource);
% DoA_error = errorDOA(theta(Ilocs),anglesTrue);
DoA_error = errorDOAcutoff(theta(Ilocs),anglesTrue,errCut);
disp(['RMSE CBF : ',num2str(sqrt(mean(power(DoA_error,2))))])
if nsim==1 && xaxis==1, outputsCBF = []; end
outputCBF = struct('theta',theta(Ilocs),'error',DoA_error);
outputsCBF = [outputsCBF; outputCBF];
%% root-MUSIC, "NUA", irregular root-MUSIC
[t_est,~] = wagner_decomp( q, Nsource, Ryy );
t_est = -t_est*lambda/d;
DoA_est_deg = asin(t_est)/pi*180;
DoA_error = errorDOAcutoff(DoA_est_deg,anglesTrue,errCut);
disp(['RMSE irr. root-MUSIC: ',num2str(sqrt(mean(power(DoA_error,2))))])
% %for ULA
% spRmusic = rmusic_1d(Ryy, Nsource, 2*pi*d/lambda);
% DoA_error = errorDOAcutoff(-rad2deg(spRmusic.x_est),anglesTrue,errCut);
% disp(['RMSE root-MUSIC: ',num2str(sqrt(mean(power(DoA_error,2))))])
if nsim==1 && xaxis==1, outputsrMUSIC = []; end
% outputrMUSIC = struct('theta',-rad2deg(spRmusic.x_est),'error',DoA_error);
outputrMUSIC = struct('theta',DoA_est_deg,'error',DoA_error);
outputsrMUSIC = [outputsrMUSIC; outputrMUSIC];
%% AP-ULA, "NUA"
% MAX_IT = 1000;
% NeigAP = Nsource;
% tol = 1e-3;
%
% [T,iAPM] = AP_Gridless( receivedSignal,q,NeigAP,MAX_IT,tol,zeros(Nsensor),0 );
% [t_est,~] = wagner_decomp( q, NeigAP, T ); %decompose
% t_est = -t_est*lambda/d;
% DoA_est_deg = asin(t_est)/pi*180;
% DoA_error = errorDOAcutoff(DoA_est_deg,anglesTrue,errCut);
% disp(['RMSE AP-Snapshot ULA: ',num2str(sqrt(mean(power(DoA_error,2))))])
%
% if nsim==1 && xaxis==1, outputsAPula = []; end
% outputAPula = struct('theta',DoA_est_deg,'error',DoA_error);
% outputsAPula = [outputsAPula; outputAPula];
%% AP-Gridless for NUA
MAX_IT = 1000;
NeigAP = Nsource;
tol = 1e-3;
% [T,iAPM] = AP_Gridless( receivedSignal,q,NeigAP,MAX_IT,tol,zeros(Nsensor),0 ); % for AP-ULA
[T,iAPM] = AP_Gridless( receivedSignal,q+.1,NeigAP,MAX_IT,tol,zeros(Nsensor),0 ); % for AP-Gridless
[t_est,~] = wagner_decomp( q, NeigAP, T ); %decompose
t_est = -t_est*lambda/d;
DoA_est_deg = asin(t_est)/pi*180;
DoA_error = errorDOAcutoff(DoA_est_deg,anglesTrue,errCut);
disp(['RMSE AP-Snapshot : ',num2str(sqrt(mean(power(DoA_error,2))))])
if nsim==1 && xaxis==1, outputsAPsnapshot = []; end
outputAPsnapshot = struct('theta',DoA_est_deg,'error',DoA_error);
outputsAPsnapshot = [outputsAPsnapshot; outputAPsnapshot];
%% AP-Covariance
max_iter = 1000;
Nalg = Nsource;
tol = 1e-4;
if exist('Tu_init','var') == 0
Tu_init = rand(Nsensor) + 1i*rand(Nsensor);
Z_init = rand(Nsensor) + 1i*rand(Nsensor);
end
[ Tu,iAP,~,~,~ ] = APCOVv1p00(receivedSignal,q,Nalg,max_iter,tol,Tu_init,Z_init);
[t_est,~] = wagner_decomp( q, Nalg, Tu ); %decompose
t_est = -t_est*lambda/d;
while(1)
t_est(t_est>1) = t_est(t_est>1) - 2;
t_est(t_est<-1)= t_est(t_est<-1) + 2;
if sum(t_est>1 | t_est<-1) == 0, break; end
end
DoA_est_deg = asin(t_est)/pi*180;
clear Tu_init Z_init
DoA_error = errorDOAcutoff(DoA_est_deg,anglesTrue,errCut);
disp(['RMSE AP-Covariance : ',num2str(sqrt(mean(power(DoA_error,2))))])
if nsim==1 && xaxis==1, outputsAPcov = []; end
outputAPcov = struct('theta',DoA_est_deg,'error',DoA_error);
outputsAPcov = [outputsAPcov; outputAPcov];
%% SBL
options = SBLSet();
options.convergence.error = 10^(-3);
options.Nsource = ceil(Nsensor/2);
options.gamma_range=10^-20;
[gamma, reportSBL] = SBL_v4( sensingMatrix, receivedSignal, options );
[~,peak_SBL] = findpeaks(gamma,'SORTSTR','descend','Npeaks', Nsource);
DoA_error = errorDOAcutoff(theta(peak_SBL),anglesTrue,errCut);
disp(['RMSE SBL : ',num2str(sqrt(mean(power(DoA_error,2))))])
if nsim==1 && xaxis==1, outputsSBL = []; end
outputSBL = struct('theta',theta(peak_SBL),'error',DoA_error);
outputsSBL = [outputsSBL; outputSBL];
end
end
%% Figure
for ind=1:length(xAxes)
totETcbf = [];
totETrmu=[];
% totETAPula=[];
totETAPsnapshot=[];
totETAPcov=[];
totETsbl = [];
for index=1:Nsim
totETcbf = [totETcbf;outputsCBF((ind-1)*Nsim+index).error];
totETrmu = [totETrmu;outputsrMUSIC((ind-1)*Nsim+index).error];
% totETAPula = [totETAPula;outputsAPula((ind-1)*Nsim+index).error];
totETAPsnapshot = [totETAPsnapshot;outputsAPsnapshot((ind-1)*Nsim+index).error];
totETAPcov = [totETAPcov;outputsAPcov((ind-1)*Nsim+index).error];
totETsbl = [totETsbl;outputsSBL((ind-1)*Nsim+index).error];
end
Nout = 0; % Portion of Outliers, (ignore)
totETcbf = sort(abs(totETcbf));
totETrmu = sort(abs(totETrmu));
% totETAPula = sort(abs(totETAPula));
totETAPsnapshot = sort(abs(totETAPsnapshot));
totETAPcov = sort(abs(totETAPcov));
totETsbl = sort(abs(totETsbl));
ecbf(ind) = sqrt(mean(power(totETcbf(1:length(totETcbf)-floor(length(totETcbf)*Nout)),2)));
ermusic(ind) = sqrt(mean(power(totETrmu(1:length(totETrmu)-floor(length(totETrmu)*Nout)),2)));
% eapula(ind) = sqrt(mean(power(totETAPula(1:length(totETAPula)-floor(length(totETAPula)*Nout)),2)));
eapsnapshot(ind) = sqrt(mean(power(totETAPsnapshot(1:length(totETAPsnapshot)-floor(length(totETAPsnapshot)*Nout)),2)));
eapcov(ind) = sqrt(mean(power(totETAPcov(1:length(totETAPcov)-floor(length(totETAPcov)*Nout)),2)));
esbl(ind) = sqrt(mean(power(totETsbl(1:length(totETsbl)-floor(length(totETsbl)*Nout)),2)));
end
figure; set(gcf,'position',[750,200,700,600]);
hold on;
h1=plot(xAxes,sqrt(mean(outputsCRBd,2)*180/pi*180/pi),'k','linewidth',1.0,'markersize',10,'displayname','det. CRB');
figH = h1;
h3=plot(xAxes,ecbf,'k:','linewidth',1.5,'markersize',10,'displayname','CBF');
figH = [figH,h3];
pcolor = lines;
h5=plot(xAxes,esbl,'m--','linewidth',1.2,'markersize',10,'displayname',...
'SBL','color',pcolor(2,:)); figH = [figH,h5];
h8=plot(xAxes,ermusic,'b-.','linewidth',2,'markersize',10,'displayname',...
'irr. Root-MUSIC'); figH = [figH,h8];
% h9=plot(xAxes,eapula,'c-.','linewidth',1.8,'markersize',12,'displayname',...
% 'AP-ULA'); figH = [figH,h9];
h91=plot(xAxes,eapsnapshot,'linewidth',1.8,'markersize',12,'color',pcolor(7,:),'displayname',...
'AP-Snapshot'); figH = [figH,h91];
h92=plot(xAxes,eapcov,'linewidth',1.8,'markersize',12,'color','r','displayname',...
'AP-Covariance'); figH = [figH,h92];
hold off;
xlabel('SNR~[dB]','interpreter','latex')
ylabel('RMSE~[$^\circ$]','interpreter','latex')
legend(fliplr(figH),'location','northeast','interpreter','latex')
set(gca,'fontsize',18,'yscale','log')
box on; grid on;
axis([min(xAxes) max(xAxes) min(sqrt(mean(outputsCRBd,2)*180/pi*180/pi)) 1])
% axis([0 20 0 10])
%%
rmpath([cd,'/_common'])