def daily_motion_test(db,avaliable_sensor):
time_interval = ['2016-12-07 13:00:00', '2016-12-07 14:00:00']
if avaliable_sensor['kinect']:
##get data from database
kinect_joints = database.read_kinect_joints_from_db(db.Kinect,time_interval)
#extract features from kinect
hours = 0
minutes = 0
seconds = 25
time_slice_size = [hours, minutes, seconds]
global_traj_features = kinect_global_trajectories.feature_extraction_video_traj(kinect_joints,time_slice_size)
#read model from database
model= database.get_classifier_model(filename='startPeriod_endPeriod')
#test classification
classifiers.logistic_regression_predict(global_traj_features[1])
##visulaization daily motion
visualization.bar_plot_occupancy_selectedAreas_over_time(global_traj_features[0])
visualization.bar_plot_motion_over_time(global_traj_features[1])
visualization.pie_plot_motion_day(global_traj_features[1])
if avaliable_sensor['zenith']:
print 'add img processing zenith camera'
zenith_data = database.read_zenith_from_db(db.Zenith,time_interval)
if avaliable_sensor['UPMBand']:
print 'add upm band processing'
upmBand_data = database.read_UPMBand_from_db(db.UPMBand,time_interval)
def daily_motion_training(db,avaliable_sensor):
time_interval = ['2017-07-12 08:03:00', '2017-07-12 08:05:00']
if avaliable_sensor['kinect']:
print 'kinect for daily motion'
##get data from database
kinect_joints = database.read_kinect_joints_from_db(db.Kinect,time_interval,multithread=0)
#extract features from kinect
hours = 0
minutes = 0
seconds = 4
date = '2017-07-11'
time_slice_size = [hours, minutes, seconds]
global_traj_features,patient_ID = kinect_global_trajectories.feature_extraction_video_traj(kinect_joints,time_slice_size, draw_joints_in_scene=1)
#Get labels from clustering
#labels_cluster = classifiers.cluster_kmeans(global_traj_features[1],k=3)
#Train in supervised way a classifier with extracted features and labels
#model = classifiers.logistic_regression_train(global_traj_features[1],labels_cluster)
#Save model in database
#database.save_classifier_model(model,filename='startPeriod_endPeriod')
##visulaization daily motion
#visualization.bar_plot_occupancy_selectedAreas_over_time(global_traj_features[0])
#visualization.bar_plot_motion_over_time(global_traj_features[1])
kinect_motion_amount = visualization.pie_plot_motion_day(global_traj_features[1])
##make it dictionary
kinect_motion_amount = {'stationary': kinect_motion_amount[0][0],'slow_mov': kinect_motion_amount[0][1],'fast_mov': kinect_motion_amount[0][2]}
if avaliable_sensor['zenith']:
print 'add img processing zenith camera'
zenith_data = database.read_zenith_from_db(db.Zenith,time_interval)
if avaliable_sensor['UPMBand']:
print 'add upm band processing'
upmBand_data = database.read_UPMBand_from_db(db.UPMBand,time_interval)
return kinect_motion_amount,patient_ID
def abnormal_behavior_classification_training(db, avaliable_sensor):
##get data in the selected time interval from database
time_interval = ['2016-12-07 13:08:00', '2016-12-07 13:15:00']
if avaliable_sensor['kinect']:
print 'kinect disorientation tr'
##get data from database
kinect_joints = database.read_kinect_joints_from_db(db.Kinect, time_interval, multithread=0)
# extract features from kinect
hours = 0
minutes = 0
seconds = 25
time_slice_size = [hours,minutes,seconds]
global_traj_features = kinect_global_trajectories.feature_extraction_video_traj(kinect_joints,time_slice_size)
## perform clustering
#classifiers.cluster_meanShift(global_traj_features[1],save_model=1)
##retrieve model and get labels
#cluster_model = database.get_classifier_model('')
##get labels from clustering
#labels_cluster = cluster_model.predict(global_traj_features[1])
##Create bag of words with trajectories
#bow_vocabulary = kinect_global_trajectories.bow_traj(global_traj_features[1],cluster_model,labels_cluster)
#classifiers.logistic_regression_train(bow_vocabulary,labels_cluster,save_model=1)
if avaliable_sensor['zenith']:
zenith_data = database.read_zenith_from_db(db.Zenith,time_interval)
if avaliable_sensor['UPMBand']:
upmBand_data = database.read_UPMBand_from_db(db.UPMBand,time_interval)
def daily_motion_training(db,avaliable_sensor, time_interval):
if avaliable_sensor['kinect']:
print 'kinect for daily motion'
##get data from database
kinect_joints = database.read_kinect_joints_from_db(db.Kinect,time_interval)
#kinect_joints = kinect_joints[::-1]
bands_ids = database.get_bands_ID(db)
if len(kinect_joints) <1:
print '----- no data in time interval -----'
kinect_motion_amount_band = {b[0]: None for b in bands_ids}
for b in bands_ids: kinect_motion_amount_band[b[0]] = {'stationary': -1,'slow_mov': -1,'fast_mov': -1}
return kinect_motion_amount_band
#extract HOT features from kinect from each user
global_traj_features = kinect_features.feature_extraction_video_traj(kinect_joints, bands_ids, draw_joints_in_scene=0, realtime=0)
kinect_motion_amount_band = {b[0]: None for b in bands_ids}
## compute analysis per band
for i_b,b in enumerate(bands_ids):
if len(global_traj_features[1][i_b])>0:
#Get labels from clustering
#labels_cluster = classifiers.cluster_kmeans(global_traj_features[1],k=3)
#Train in supervised way a classifier with extracted features and labels
#model = classifiers.logistic_regression_train(global_traj_features[1],labels_cluster)
#Save model in database
#database.save_classifier_model(model,filename='startPeriod_endPeriod')
##visulaization daily motion
#visualization.bar_plot_occupancy_selectedAreas_over_time(global_traj_features[0])
#visualization.bar_plot_motion_over_time(global_traj_features[1])
kinect_motion_amount = visualization.pie_plot_motion_day(global_traj_features[1][i_b],plot=0)
# normalize motion
total_motion = np.sum(kinect_motion_amount[0])
for n_k, k_data in enumerate(kinect_motion_amount[0]): kinect_motion_amount[0][n_k] = k_data / total_motion
##make it dictionary
kinect_motion_amount_band[b[0]]= {'stationary': kinect_motion_amount[0][0],'slow_mov': kinect_motion_amount[0][1],'fast_mov': kinect_motion_amount[0][2]}
print b[0], {'stationary': kinect_motion_amount[0][0],'slow_mov': kinect_motion_amount[0][1],'fast_mov': kinect_motion_amount[0][2]}
else:
print 'no data from band: ', b[0]
kinect_motion_amount_band[b[0]] = {'stationary': -1,'slow_mov': -1,'fast_mov': -1}
if avaliable_sensor['zenith']:
print 'add img processing zenith camera'
zenith_data = database.read_zenith_from_db(db.Zenith,time_interval)
if avaliable_sensor['UPMBand']:
print 'add upm band processing'
upmBand_data = database.read_UPMBand_from_db(db.UPMBand,time_interval)
return kinect_motion_amount_band