wolffd@0: classdef MTTTagFeatureGenreBasic < MTTAudioFeature & handle
wolffd@0:     
wolffd@0:  
wolffd@0:     properties(Constant = true)
wolffd@0:         
wolffd@0:         my_revision = str2double(substr('$Rev: 99 $',  5, -1));
wolffd@0:     end
wolffd@0:        
wolffd@0:     % ---
wolffd@0:     % here come the internal clip properties.
wolffd@0:     % the database is stored as a global variable
wolffd@0:     % ---
wolffd@0:     properties
wolffd@0:         
wolffd@0:        my_params = struct(...
wolffd@0:             ... % ---
wolffd@0:             ... % these are GenreBasic parameters
wolffd@0:             ... % ---
wolffd@0:             'pct_genres', 1, ... % 1/100 percentile genre tags used
wolffd@0:             'empty_genres', 0 ... % allow empty genres to persist
wolffd@0:             );
wolffd@0:     end
wolffd@0:     
wolffd@0:     methods
wolffd@0:         
wolffd@0:         % ---
wolffd@0:         % constructor: pointer to feature in database
wolffd@0:         % ---
wolffd@0:         function feature = MTTTagFeatureGenreBasic(varargin)
wolffd@0: 
wolffd@0:             feature = feature@MTTAudioFeature(varargin{:});
wolffd@0:         end
wolffd@0:  
wolffd@0:         function data = extract(feature, clip)
wolffd@0:         % process tag information and build the tag vector
wolffd@0:             
wolffd@0:             % ---
wolffd@0:             % get total number of tags. 
wolffd@0:             %
wolffd@0:             % NOTE: the highest tag id should match the size of the 
wolffd@0:             % tagdb lexicon
wolffd@0:             % ---
wolffd@0: %             num_tags = clip.my_db.genredb.size();
wolffd@0:             
wolffd@0:             if feature.my_params.pct_genres ~= 0
wolffd@0:                 % no tags, f.e. for parameter experiments in
wolffd@0:                 % higher level features
wolffd@0:                 [tagids, score] = clip.my_db.genredb.annotids_for_owner(clip.id);
wolffd@0: 
wolffd@0:                 % save lexicon
wolffd@0:                 data.lexicon = clip.my_db.genredb.lexicon;
wolffd@0: 
wolffd@0:                 % save to data structure
wolffd@0:                 data.tags.ids = tagids;
wolffd@0:                 data.tags.scores = score;
wolffd@0:             end
wolffd@0:             
wolffd@0:             % save info data
wolffd@0:             data.info.type = 'MTTTagFeatureGenreBasic';
wolffd@0:             data.info.owner = clip;
wolffd@0:             data.info.owner_id = clip.id;
wolffd@0:             data.info.creatorrev = feature.my_revision;
wolffd@0:             
wolffd@0:             % save parameters
wolffd@0:             data.info.params = feature.my_params;
wolffd@0:             
wolffd@0:             data.final.vector = [];
wolffd@0:             data.final.vector_info = struct(); 
wolffd@0:             data.final.dim = 0;
wolffd@0:         end
wolffd@0:         
wolffd@0:         
wolffd@0:         % ---
wolffd@0:         % NOTE: the following transforms are just stated fro 
wolffd@0:         % future use by now 
wolffd@0:         % ---
wolffd@0:         function define_global_transform(features)
wolffd@0:             % ---
wolffd@0:             % compute the relevant tags, and save
wolffd@0:             % them in the commom place
wolffd@0:             % ---
wolffd@0:             
wolffd@0:             % compute extreme cases
wolffd@0:             if features(1).my_params.pct_genres == 1 && features(1).my_params.empty_genres
wolffd@0:                 % all tags allowed
wolffd@0:                 
wolffd@0:                 common.rel_dimensions.tags.ids = 1:numel(features(1).data.lexicon);
wolffd@0:                 common.rel_dimensions.tags.id_pos = 1:numel(features(1).data.lexicon);
wolffd@0:                 
wolffd@0:                 % set common feature values
wolffd@0:                 features(1).my_db.set_common(common);
wolffd@0:                 return;
wolffd@0:                 
wolffd@0:             elseif features(1).my_params.pct_genres == 0 
wolffd@0:                 % no tags, f.e. for parameter experiments in
wolffd@0:                 % higher level features
wolffd@0:                 
wolffd@0:                 common.rel_dimensions.tags.ids = [];
wolffd@0:                 common.rel_dimensions.tags.id_pos = [];
wolffd@0:                 
wolffd@0:                 % set common feature values
wolffd@0:                 features(1).my_db.set_common(common);
wolffd@0:                 return;
wolffd@0:             end
wolffd@0:             
wolffd@0:             
wolffd@0:             allids = sparse(numel(features), numel(features(1).data.lexicon));
wolffd@0:             for i = 1:numel(features)
wolffd@0:                 
wolffd@0:                 allids(i,features(i).data.tags.ids) = ...
wolffd@0:                     allids(i,features(i).data.tags.ids) + features(i).data.tags.scores;
wolffd@0:             end
wolffd@0:             
wolffd@0:             % ---
wolffd@0:             % get usage of tags and filter not-used tags
wolffd@0:             % ---
wolffd@0:             tagsum = sum(allids, 1);
wolffd@0:             nonzero = tagsum > 0;
wolffd@0:             
wolffd@0:             % ---
wolffd@0:             % NOTE: We remove the empty genres, then
wolffd@0:             % sort by weight / number of appearance and cut off
wolffd@0:             ids = find(nonzero);
wolffd@0:             [null, idx] = sort(tagsum(ids),'descend');
wolffd@0:             ids = ids(idx);
wolffd@0:             
wolffd@0:             % cutoff
wolffd@0:             num_tags = ceil( features(1).my_params.pct_genres * numel(ids));
wolffd@0:             valid_ids = ids(1:min(end, num_tags));
wolffd@0:             
wolffd@0:             % ---
wolffd@0:             % NOTE: make sure that the positions for the tags
wolffd@0:             % stay correctly assigned / consistent
wolffd@0:             % ---
wolffd@0:             id_pos = sparse(size(valid_ids));
wolffd@0:             id_pos(valid_ids) = 1:num_tags; 
wolffd@0:             
wolffd@0:             % save to common data structure
wolffd@0:             common.rel_dimensions.tags.ids = valid_ids;
wolffd@0:             common.rel_dimensions.tags.id_pos = id_pos;
wolffd@0:      
wolffd@0:             % set common feature values
wolffd@0:             features(1).my_db.set_common(common);
wolffd@0:         end
wolffd@0:         
wolffd@0:         function finalise(features)
wolffd@0:           
wolffd@0:             for i = 1:numel(features)
wolffd@0:                 
wolffd@0:                  % check for neccesary parameters
wolffd@0:                 if isempty(features(i).my_db.commondb)
wolffd@0: 
wolffd@0:                     error('Define the global transformation first')
wolffd@0:                     return;
wolffd@0:                 end
wolffd@0:                 
wolffd@0:                 if features(i).my_params.pct_genres == 0
wolffd@0:                     vector_info.labels = {};
wolffd@0: 
wolffd@0:                     % save into feature struct
wolffd@0:                     features(i).data.final.dim = 0;
wolffd@0:                     features(i).data.final.vector = [];
wolffd@0:                     features(i).data.final.vector_info = vector_info;
wolffd@0:                     return;
wolffd@0:                 end
wolffd@0:                 % get valid tag ids
wolffd@0:                 valid_ids = features(i).common.rel_dimensions.tags.ids;
wolffd@0:                 num_tags = numel(valid_ids);
wolffd@0:                 
wolffd@0:                 % ---
wolffd@0:                 % get vector positions for contined ids,
wolffd@0:                 % and sort out the non_allowed id's
wolffd@0:                 % ---
wolffd@0:                 id_pos = features(i).common.rel_dimensions.tags.id_pos;
wolffd@0:                 
wolffd@0:                 [has_ids, has_pos] = intersect(features(i).data.tags.ids, valid_ids);
wolffd@0:                 
wolffd@0:                 score = features(i).data.tags.scores;
wolffd@0:                 
wolffd@0:                 % create feature vector
wolffd@0:                 vector = zeros(num_tags,1);     
wolffd@0:                 vector(id_pos(has_ids)) = score(has_pos);
wolffd@0:                 
wolffd@0:                 % ---
wolffd@0:                 % NOTE: this labelling costs to much space
wolffd@0:                 % ---
wolffd@0:                 vector_info.labels = features(i).data.lexicon(valid_ids);
wolffd@0: 
wolffd@0:                 % save into feature struct
wolffd@0:                 features(i).data.final.dim = num_tags;
wolffd@0:                 features(i).data.final.vector = vector;
wolffd@0:                 features(i).data.final.vector_info = vector_info;
wolffd@0:             end
wolffd@0:         end
wolffd@0:         
wolffd@0:         function [a1] = visualise(feature)
wolffd@0:             
wolffd@0:             % get tag descriptions form first clip
wolffd@0:             tags = MTTClip(feature(1).owner_id()).my_db.genres;
wolffd@0:             
wolffd@0:             % works for multiple feature instances
wolffd@0:              for i = 1:numel(feature)
wolffd@0:                   
wolffd@0:                 clip = MTTClip(feature(i).owner_id());
wolffd@0:                 
wolffd@0:                 % plot feature data
wolffd@0:                 h = bar(feature(i).data.final.vector);
wolffd@0:                 a1 = gca;
wolffd@0:                 
wolffd@0:                 set(a1,'XTick', 1:feature(i).data.final.dim,...
wolffd@0:                     'XTickLabel', tags);
wolffd@0:                 title(sprintf('clip %d: %s by %s, genres', ...
wolffd@0:                     clip.id, clip.title(),clip.artist()));
wolffd@0:              end
wolffd@0:         end
wolffd@0: 
wolffd@0:     end
wolffd@0: end