Mercurial > hg > ltpda

comparison m-toolbox/classes/@ao/lpsd.m @ 0:f0afece42f48

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Import.
author Daniele Nicolodi <nicolodi@science.unitn.it>
date Wed, 23 Nov 2011 19:22:13 +0100
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:f0afece42f48
1 % LPSD implements the LPSD algorithm for analysis objects.
2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3 %
4 % DESCRIPTION: LPSD implements the LPSD algorithm for analysis objects.
5 %
6 % CALL: bs = lpsd(a1,a2,a3,...,pl)
7 % bs = lpsd(as,pl)
8 % bs = as.lpsd(pl)
9 %
10 % INPUTS: aN - input analysis objects
11 % as - input analysis objects array
12 % pl - input parameter list
13 %
14 % OUTPUTS: bs - array of analysis objects, one for each input
15 %
16 % <a href="matlab:utils.helper.displayMethodInfo('ao', 'lpsd')">Parameters Description</a>
17 %
18 % VERSION: $Id: lpsd.m,v 1.55 2011/05/22 21:22:09 mauro Exp $
19 %
20 % References: "Improved spectrum estimation from digitized time series
21 % on a logarithmic frequency axis", Michael Troebs, Gerhard Heinzel,
22 % Measurement 39 (2006) 120-129.
23 %
24 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
25
26 function varargout = lpsd(varargin)
27
28 % Check if this is a call for parameters
29 if utils.helper.isinfocall(varargin{:})
30 varargout{1} = getInfo(varargin{3});
31 return
32 end
33
34 import utils.const.*
35 utils.helper.msg(msg.PROC3, 'running %s/%s', mfilename('class'), mfilename);
36
37 % Collect input variable names
38 in_names = cell(size(varargin));
39 for ii = 1:nargin,in_names{ii} = inputname(ii);end
40
41 % Collect all AOs
42 [as, ao_invars] = utils.helper.collect_objects(varargin(:), 'ao', in_names);
43
44 % Decide on a deep copy or a modify
45 bs = copy(as, nargout);
46
47 % Apply defaults to plist
48 pl = applyDefaults(getDefaultPlist, varargin{:});
49
50 inhists = [];
51
52 % Loop over input AOs
53 for jj = 1 : numel(bs)
54 % gather the input history objects
55 inhists = [inhists bs(jj).hist];
56
57 % check this is a time-series object
58 if ~isa(bs(jj).data, 'tsdata')
59 warning('!!! lpsd requires tsdata (time-series) inputs. Skipping AO %s', ao_invars{jj});
60 else
61
62 % Check the time range.
63 time_range = mfind(pl, 'split', 'times');
64 if ~isempty(time_range)
65 switch class(time_range)
66 case 'double'
67 bs(jj) = split(bs(jj), plist(...
68 'times', time_range));
69 case 'timespan'
70 bs(jj) = split(bs(jj), plist(...
71 'timespan', time_range));
72 case 'time'
73 bs(jj) = split(bs(jj), plist(...
74 'start_time', time_range(1), ...
75 'end_time', time_range(2)));
76 case 'cell'
77 bs(jj) = split(bs(jj), plist(...
78 'start_time', time_range{1}, ...
79 'end_time', time_range{2}));
80 otherwise
81 end
82 end
83
84 % Check the length of the object
85 if bs(jj).len <= 0
86 error('### The object is empty! Please revise your settings ...');
87 end
88
89 pl = utils.helper.process_spectral_options(pl, 'log');
90
91 % Desired number of averages
92 Kdes = find(pl, 'Kdes');
93 % num desired spectral frequencies
94 Jdes = find(pl, 'Jdes');
95 % Minimum segment length
96 Lmin = find(pl, 'Lmin');
97 % Window function
98 Win = find(pl, 'Win');
99 % Overlap
100 Nolap = find(pl, 'Olap')/100;
101 % Order of detrending
102 Order = find(pl, 'Order');
103
104 % Get frequency vector
105 [f, r, m, L, K] = ao.ltf_plan(length(bs(jj).data.y), bs(jj).data.fs, Nolap, 1, Lmin, Jdes, Kdes);
106
107 % compute LPSD
108 try
109 if find(pl, 'M-FILE ONLY')
110 % Using pure m-file version
111 [P, Pxx, ENBW] = ao.mlpsd_m(bs(jj).data.y, f, r, m, L, bs(jj).data.fs, Win, Order, Nolap);
112 else
113 [P, Pxx, dev, devxx, ENBW] = ao.mlpsd_mex(bs(jj).data.y, f, r, m, L, bs(jj).data.fs, Win, Order, Nolap*100, Lmin);
114 end
115 catch ME
116 warning('!!! mex file dft failed. Using m-file version of lpsd.');
117 % Using pure m-file version
118 [P, Pxx, ENBW] = ao.mlpsd_m(bs(jj).data.y, f, r, m, L, bs(jj).data.fs, Win, Order, Nolap);
119 end
120
121 % Keep the data shape of the input AO
122 if size(bs(jj).data.y,1) == 1
123 P = P.';
124 Pxx = Pxx.';
125 dev = dev.';
126 devxx = devxx.';
127 f = f.';
128 end
129
130 % create new output fsdata
131 scale = find(pl, 'Scale');
132 switch lower(scale)
133 case 'as'
134 fsd = fsdata(f, sqrt(P), bs(jj).data.fs);
135 fsd.setYunits(bs(jj).data.yunits);
136 std = sqrt(dev);
137 case 'asd'
138 fsd = fsdata(f, sqrt(Pxx), bs(jj).data.fs);
139 fsd.setYunits(bs(jj).data.yunits / unit('Hz^0.5'));
140 std = sqrt(devxx);
141 case 'ps'
142 fsd = fsdata(f, P, bs(jj).data.fs);
143 fsd.setYunits(bs(jj).data.yunits.^2);
144 std = dev;
145 case 'psd'
146 fsd = fsdata(f, Pxx, bs(jj).data.fs);
147 fsd.setYunits(bs(jj).data.yunits.^2/unit('Hz'));
148 std = devxx;
149 otherwise
150 error(['### Unknown scaling:' scale]);
151 end
152 fsd.setXunits('Hz');
153 fsd.setEnbw(ENBW);
154 fsd.setT0(bs(jj).data.t0);
155 % make output analysis object
156 bs(jj).data = fsd;
157 % set name
158 bs(jj).name = sprintf('L%s(%s)', upper(scale), ao_invars{jj});
159 % Add processing info
160 bs(jj).procinfo = plist('r', r, 'm', m, 'l', L, 'k', K);
161 % Add standard deviation
162 bs(jj).data.dy = std;
163 % Add history
164 bs(jj).addHistory(getInfo('None'), pl, ao_invars(jj), inhists(jj));
165
166 end % End tsdata if/else
167 end % loop over analysis objects
168
169 % Set output
170 varargout = utils.helper.setoutputs(nargout, bs);
171 end
172
173 %--------------------------------------------------------------------------
174 % Get Info Object
175 %--------------------------------------------------------------------------
176 function ii = getInfo(varargin)
177 if nargin == 1 && strcmpi(varargin{1}, 'None')
178 sets = {};
179 pl = [];
180 else
181 sets = {'Default'};
182 pl = getDefaultPlist();
183 end
184 % Build info object
185 ii = minfo(mfilename, 'ao', 'ltpda', utils.const.categories.sigproc, '$Id: lpsd.m,v 1.55 2011/05/22 21:22:09 mauro Exp $', sets, pl);
186 end
187
188 %--------------------------------------------------------------------------
189 % Get Default Plist
190 %--------------------------------------------------------------------------
191 function plout = getDefaultPlist()
192 persistent pl;
193 if ~exist('pl', 'var') || isempty(pl)
194 pl = buildplist();
195 end
196 plout = pl;
197 end
198
199 function pl = buildplist()
200
201 % General plist for Welch-based, log-scale spaced spectral estimators
202 pl = plist.LPSD_PLIST;
203
204 % Scale
205 p = param({'Scale',['The scaling of output. Choose from:<ul>', ...
206 '<li>PSD - Power Spectral Density</li>', ...
207 '<li>ASD - Amplitude (linear) Spectral Density</li>', ...
208 '<li>PS - Power Spectrum</li>', ...
209 '<li>AS - Amplitude (linear) Spectrum</li></ul>']}, {1, {'PSD', 'ASD', 'PS', 'AS'}, paramValue.SINGLE});
210 pl.append(p);
211
212 end
213
214 % PARAMETERS:
215 %
216 % 'Kdes' - desired number of averages to perform [default: 100]
217 % 'Jdes' - number of spectral frequencies to compute [default: 1000]
218 % 'Lmin' - minimum segment length [default: 0]
219 % 'Win' - the window to be applied to the data to remove the
220 % discontinuities at edges of segments. [default: taken from
221 % user prefs]
222 % Only the design parameters of the window object are
223 % used. Enter either:
224 % - a specwin window object OR
225 % - a string value containing the window name
226 % e.g., plist('Win', 'Kaiser', 'psll', 200)
227 % 'Olap' - segment percent overlap [default: -1, (taken from window function)]
228 % 'Scale' - scaling of output. Choose from:
229 % PSD - Power Spectral Density [default]
230 % ASD - Amplitude (linear) Spectral Density
231 % PS - Power Spectrum
232 % AS - Amplitude (linear) Spectrum
233 % 'Order' - order of segment detrending
234 % -1 - no detrending
235 % 0 - subtract mean [default]
236 % 1 - subtract linear fit
237 % N - subtract fit of polynomial, order N