def feature_extractor_one_dim(data): feature_vector = [] c = 0 for j in range(0,len(data)): feature_vector.append([]) for w in range(len(ref_samples)): for d in ref_indexes[w]: feature_vector[c].append(arc.my_kernel(arc, data[j][d], ref_samples[w][d], learning_rate=0.01)) c = c + 1 return feature_vector
from myfunctions import Action_recognition_class as arc from sklearn.svm import SVC import random import numpy as np #reading sata all_actions, labels = arc.read_UTkinect_aligned() #preparing data all_action_prepared = arc.zero_padding(arc, all_actions) all_action_prepared = [x[3:] for x in all_action_prepared] num_of_classes = len(set(labels)) #============================================================================================= ref_samples = [] for i in range(0, num_of_classes): while (len(ref_samples) < (i + 1)): j = random.randint(0, len(all_action_prepared) - 1) if (labels[j] == i): ref_samples.append(all_action_prepared[j]) #============================================================================================= ref_indexes = [x for x in range(len(ref_samples[0]))] ref_indexes = [ref_indexes for i in range(num_of_classes)] ''' two functions that extract feature first one is based on multi dimentional kernel second one is based on one dimentional kernel ''' '''
from myfunctions import Action_recognition_class as arc from sklearn.svm import SVC import random import numpy as np #reading TST fall detection data all_actions, labels = arc.read_tst() #============================================================================================= #Preparing data all_action_prepared = arc.prepare_data_tst(arc, all_actions) all_action_prepared = arc.zero_padding(arc, all_actions) #some samples should be deleted because of different series length of features del all_action_prepared[78] del all_action_prepared[219] del all_action_prepared[234] del labels[78] del labels[219] del labels[234] num_of_classes = len(set(labels)) #============================================================================================= #choosing reference samples randomly ref_samples = [] for i in range(0,num_of_classes): while(len(ref_samples) < (i + 1)): j = random.randint(0, len(all_action_prepared) - 1)