find_means.rar_This Is It

% FIND_MEANS Find means of class features % % Given a set of training data features, we wish to find the mean of each feature % for a given class of character. These mean values are then normalized to permit % accurate distance measurements. function T = find_means(stats,feature_name) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Create a sorted list of the character classes num = length(stats); classes(1) = stats(1).Class; ind = 1; for k = 2:num % if the class has not yet been encountered if ~sum(classes==stats(k).Class) ind = ind + 1; classes(ind,:) = stats(k).Class; end end classes = sort(classes); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Create a matrix of mean feature values for each class of character cc = [stats(:).Class]'; for feature_num = 1:length(feature_name) s = ['[stats.' feature_name{feature_num} ']''']; this_feature = eval(s); for class_num = 1:length(classes) % find which of the training characters belong to each class char_ind = cc==classes(class_num); f_temp = this_feature(char_ind); f_mean(class_num,feature_num) = sum(f_temp)/length(f_temp); end end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Normalize the class/feature means min_f_val = min(f_mean); f_mean_norm = f_mean; for feature_num = 1:length(feature_name) f_mean_norm(:,feature_num) = f_mean_norm(:,feature_num) - min_f_val(feature_num); max_f_val(feature_num) = max(f_mean_norm(:,feature_num)); f_mean_norm(:,feature_num) = f_mean_norm(:,feature_num) / max_f_val(feature_num); end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Store the important variables in a structure T.classes = classes; T.min_f_val = min_f_val; T.max_f_val = max_f_val; T.f_mean_norm = f_mean_norm;