Mercurial > hg > camir-aes2014

annotate core/magnatagatune/MTTMixedFeatureGenreBasicSmPCA.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 % --
wolffd@0 2 % This class loads and hanles the aufdio features included with the MTT
wolffd@0 3 % Library
wolffd@0 4 % ---
wolffd@0 5 classdef MTTMixedFeatureGenreBasicSmPCA < MTTAudioFeature & handle
wolffd@0 6
wolffd@0 7 properties(Constant = true)
wolffd@0 8
wolffd@0 9 my_revision = str2double(substr('$Rev$', 5, -1));
wolffd@0 10 end
wolffd@0 11
wolffd@0 12 properties
wolffd@0 13
wolffd@0 14 % ---
wolffd@0 15 % Set default parameters
wolffd@0 16 % ---
wolffd@0 17 my_basetype = 'MTTMixedFeatureGenreBasicSm';
wolffd@0 18
wolffd@0 19 my_params = MTTAudioFeature.inherited_params(...
wolffd@0 20 'MTTMixedFeatureGenreBasicSm', ...
wolffd@0 21 'min_pca_var', 0, ... % % fraction of variance to keep
wolffd@0 22 'max_pca_coeffs', 0, ...% max. number of final coefficients
wolffd@0 23 'norm_pre_pca', 1, ...% normalise pca coefficients after transformation
wolffd@0 24 'norm_post_pca', 1 ...
wolffd@0 25 );
wolffd@0 26 end
wolffd@0 27 % ---
wolffd@0 28 % member functions
wolffd@0 29 % ---
wolffd@0 30 methods
wolffd@0 31
wolffd@0 32 % ---
wolffd@0 33 % constructor: pointer to feature in database
wolffd@0 34 % ---
wolffd@0 35 function feature = MTTMixedFeatureGenreBasicSmPCA(varargin)
wolffd@0 36
wolffd@0 37 feature = feature@MTTAudioFeature(varargin{:});
wolffd@0 38 end
wolffd@0 39
wolffd@0 40 % ---
wolffd@0 41 % load feature data from xml file
wolffd@0 42 % ---
wolffd@0 43 function data = extract(feature, clip)
wolffd@0 44 % load feature data by parsing xml
wolffd@0 45
wolffd@0 46 global globalvars;
wolffd@0 47
wolffd@0 48 % ---
wolffd@0 49 % we extract the base features, and save
wolffd@0 50 % the pointers to these.
wolffd@0 51 % the main work is then done in the define_global_transf
wolffd@0 52 % and finalise functions.
wolffd@0 53 % ---
wolffd@0 54 data.basefeat = clip.features(feature.my_basetype,...
wolffd@0 55 feature.my_params);
wolffd@0 56
wolffd@0 57 % save info data
wolffd@0 58 data.info.type = 'MTTMixedFeatureGenreBasicSmPCA';
wolffd@0 59 data.info.owner_id = clip.id;
wolffd@0 60 data.info.creatorrev = feature.my_revision;
wolffd@0 61
wolffd@0 62 % save param data
wolffd@0 63 data.info.params = feature.my_params;
wolffd@0 64
wolffd@0 65 % prepare field for final features
wolffd@0 66 data.final.vector = [];
wolffd@0 67 data.final.dim = 0;
wolffd@0 68 data.final.vector_info.labels = {};
wolffd@0 69 end
wolffd@0 70
wolffd@0 71 function define_global_transform(features)
wolffd@0 72
wolffd@0 73 if numel(features) == 1
wolffd@0 74 error ('Insert feature array for this method');
wolffd@0 75 end
wolffd@0 76
wolffd@0 77 % ---
wolffd@0 78 % We collect all the relevant genretag
wolffd@0 79 % features and get the transform on this basis.
wolffd@0 80 % ---
wolffd@0 81 for i = 1:numel(features)
wolffd@0 82 basef(i) = features(i).data.basefeat;
wolffd@0 83 end
wolffd@0 84
wolffd@0 85 % call the features own transsform function
wolffd@0 86 basef.define_global_transform();
wolffd@0 87
wolffd@0 88 % ---
wolffd@0 89 % finalise the basic features, and
wolffd@0 90 % get the feature vectors;
wolffd@0 91 % ---
wolffd@0 92 basef.finalise();
wolffd@0 93 X = basef.vector();
wolffd@0 94
wolffd@0 95 % check dataset dimension
wolffd@0 96 if numel(features) < basef.dim;
wolffd@0 97
wolffd@0 98 error ('Not enough feature vectors for PCA calculation. need %d samples', ...
wolffd@0 99 basef.dim);
wolffd@0 100 end
wolffd@0 101
wolffd@0 102 % ---
wolffd@0 103 % NOTE: should the data be normalised and scaled to -1:1
wolffd@0 104 % instead of being in a range of 0-1 AND max-min = 1
wolffd@0 105 % ---
wolffd@0 106 if features(1).my_params.norm_pre_pca == 1
wolffd@0 107
wolffd@0 108 [X, pstd] = mapminmax(X,-1,1);
wolffd@0 109 common.pca.pre_norm = pstd;
wolffd@0 110 elseif features(1).my_params.norm_pre_pca == 2
wolffd@0 111
wolffd@0 112 [X, pstd] = mapstd(X,0,1);
wolffd@0 113 common.pca.pre_norm = pstd;
wolffd@0 114 end
wolffd@0 115
wolffd@0 116
wolffd@0 117 % ---
wolffd@0 118 % get and apply the principal component analysis
wolffd@0 119 % NOTE: the variance percentile is applied here, too.
wolffd@0 120 % ---
wolffd@0 121
wolffd@0 122 [Y, ps] = processpca(X, 0);
wolffd@0 123 common.pca.transform = ps;
wolffd@0 124
wolffd@0 125 % ---
wolffd@0 126 % get cumulative sum of variance, and decide on cutoff
wolffd@0 127 % point
wolffd@0 128 % ---
wolffd@0 129 v = cumsum(var(Y'));
wolffd@0 130 v = v / max(v);
wolffd@0 131 common.pca.transform.varpart = v;
wolffd@0 132
wolffd@0 133 if features(1).my_params.min_pca_var > 0
wolffd@0 134
wolffd@0 135 min_pca_idx = find(v >= features(1).my_params.min_pca_var,1, 'first');
wolffd@0 136
wolffd@0 137 % save into pca structure
wolffd@0 138 common.pca.transform.yrows = min_pca_idx;
wolffd@0 139
wolffd@0 140 end
wolffd@0 141
wolffd@0 142 % normalise pca values after processing
wolffd@0 143 if features(1).my_params.norm_post_pca
wolffd@0 144
wolffd@0 145 [Y,pmm] = mapminmax(Y,0,1);
wolffd@0 146 common.pca.post_norm = pmm;
wolffd@0 147
wolffd@0 148 end
wolffd@0 149
wolffd@0 150 % ---
wolffd@0 151 % set common feature values for mixed features
wolffd@0 152 % ---
wolffd@0 153 features(1).my_db.set_common(common);
wolffd@0 154
wolffd@0 155 % save the transformed features straight away!
wolffd@0 156 features.finalise(Y);
wolffd@0 157 end
wolffd@0 158
wolffd@0 159 function finalise(feature, final)
wolffd@0 160 % applies a final transformation and
wolffd@0 161 % collects the information of this feature within a single vector
wolffd@0 162 % see info for types in specific dimensions
wolffd@0 163
wolffd@0 164 % determine size
wolffd@0 165 if feature(1).my_params.max_pca_coeffs > 0
wolffd@0 166
wolffd@0 167 max_size = min(feature(1).common.pca.transform.yrows, ...
wolffd@0 168 feature(1).my_params.max_pca_coeffs);
wolffd@0 169 else
wolffd@0 170
wolffd@0 171 max_size = feature(1).common.pca.transform.yrows;
wolffd@0 172 end
wolffd@0 173
wolffd@0 174
wolffd@0 175 % prepare information
wolffd@0 176 info = {'PCA'};
wolffd@0 177 if isfield(feature(1).common.pca.transform, 'varpart')
wolffd@0 178 info(2:max_size) = num2cell(feature(1).common.pca.transform.varpart(2:max_size));
wolffd@0 179 else
wolffd@0 180 info(2:max_size) = num2cell(2:max_size);
wolffd@0 181 end
wolffd@0 182
wolffd@0 183
wolffd@0 184 % check if features have been finalised already
wolffd@0 185 if nargin == 2 && isempty(final)
wolffd@0 186
wolffd@0 187 % the final vector etc already are set to zero;
wolffd@0 188 return;
wolffd@0 189
wolffd@0 190 elseif nargin == 2 && (numel(feature) == size(final, 2))
wolffd@0 191
wolffd@0 192 for i = 1:numel(feature)
wolffd@0 193
wolffd@0 194 % save final vector and description
wolffd@0 195 feature(i).data.final.vector = final(1:max_size,i);
wolffd@0 196 feature(i).data.final.dim = max_size;
wolffd@0 197 feature(i).data.final.vector_info.labels = info;
wolffd@0 198 end
wolffd@0 199
wolffd@0 200 else
wolffd@0 201 % features have to be transformed first
wolffd@0 202 % ---
wolffd@0 203 % TODO: this code remains untested
wolffd@0 204 % ---
wolffd@0 205
wolffd@0 206 % check for neccesary parameters
wolffd@0 207 if isempty(feature(1).my_db.commondb)
wolffd@0 208
wolffd@0 209 error('Define the global transformation first')
wolffd@0 210 return;
wolffd@0 211 end
wolffd@0 212
wolffd@0 213
wolffd@0 214 for i = 1:numel(feature)
wolffd@0 215
wolffd@0 216 % check for neccesary parameters
wolffd@0 217 if isempty(feature(i).my_db.commondb)
wolffd@0 218
wolffd@0 219 error('Define the global transformation first')
wolffd@0 220 end
wolffd@0 221
wolffd@0 222 % ---
wolffd@0 223 % get feature vector and apply transformation
wolffd@0 224 % ---
wolffd@0 225 X = feature(i).data.basefeat.vector();
wolffd@0 226
wolffd@0 227 % ---
wolffd@0 228 % apply normalisation used for removing mean
wolffd@0 229 % in training data
wolffd@0 230 % ---
wolffd@0 231 if feature(1).my_params.norm_pre_pca == 1
wolffd@0 232
wolffd@0 233 X = mapminmax('apply', X, feature(1).common.pca.pre_norm);
wolffd@0 234 elseif feature(1).my_params.norm_pre_pca == 2
wolffd@0 235
wolffd@0 236 X = mapstd('apply', X, feature(1).common.pca.pre_norm);
wolffd@0 237 end
wolffd@0 238
wolffd@0 239 % apply PCA transform
wolffd@0 240 vec = processpca('apply', X, feature(1).common.pca.transform);
wolffd@0 241
wolffd@0 242 % normalise pca values after transformation
wolffd@0 243 if feature(1).my_params.norm_post_pca
wolffd@0 244
wolffd@0 245 vec = mapminmax('apply', vec,...
wolffd@0 246 feature(1).common.pca.post_norm);
wolffd@0 247 end
wolffd@0 248
wolffd@0 249 % ---
wolffd@0 250 % cut vector to final size.
wolffd@0 251 % NOTE: this should be done before
wolffd@0 252 % transformation to reduce computation time
wolffd@0 253 % ---
wolffd@0 254 vec = vec(1:max_size);
wolffd@0 255
wolffd@0 256 % save final vector and description
wolffd@0 257 feature(i).data.final.vector = vec;
wolffd@0 258 feature(i).data.final.dim = numel(vec);
wolffd@0 259 feature(i).data.final.vector_info.labels = info;
wolffd@0 260 end
wolffd@0 261 end
wolffd@0 262 end
wolffd@0 263
wolffd@0 264 end
wolffd@0 265 end