Term Project codes added

Author: arnab39

Imposter detection/Feature_extraction.py

@@ -0,0 +1,197 @@
+import pandas as pd
+from datetime import datetime
+import os 
+import numpy as np
+
+
+#The path where the data is stored
+datapath_begin='New Data Collection/'
+datapath_end='/Keyboard Database/sentence/'
+
+
+#The details of the user data we are considering 
+user_details = {0:['Lalit','14EC10025'],2:['Arnab','14EC35031'],3:['Sandeep','14EC35033'],4:['Charu','14EC35003'],1:['Koruprolu Asish','14EC10024'] }
+
+
+'''To read files and extract data from it '''
+##############################################
+
+def extract_data(username,roll_number,mood,cont=False):
+
+data2 = pd.DataFrame({'key' : [],
+ 'press_time' : [],
+ 'latency' : []})
+path = datapath_begin + username + '_' + roll_number + datapath_end + mood +'/'
+directory = os.path.join(path)
+print 'Collecting data of '+username+'...'
+for root,dirs,files in os.walk(directory):
+for file in files:
+if file.endswith(".txt"):
+with open(directory+file,"r") as file1:
+if cont:
+FMT = '%d:%H:%M:%S.%f'
+else:
+FMT = '%d:%m:%Y:%H:%M:%S.%f' 
+loop = 0
+queue = [[],[]]
+data1 = pd.DataFrame({'key' : [],
+'time' : [],
+ 'key_press_time' : []})
+for line in file1:
+loop += 1
+#refine data
+if cont:
+a = line.split()[0]
+b = line.split()[1]
+c = line.split()[7]
+
+else: 
+if len(line.split()) == 3:
+[a,b,c] = line.split()
+if b=='\b':
+b="backspace"
+c = c[:-4] + '.' + c[-3:]
+c = c[:6] + '20' + c[6:]
+
+elif len(line.split()) == 2:
+[a,c] = line.split()
+b = "space"
+c = c[:-4] + '.' + c[-3:]
+c = c[:6] + '20' + c[6:]
+
+else:
+print "Error in Data"+str(file)
+
+#record keypress
+if a == 'KeyDown':
+try:
+index_of_letter = queue[0].index(b)
+except:
+queue[0].append(b)
+queue[1].append(c)
+else:
+continue
+
+#compute data for key release
+if loop > 1 and a == 'KeyUp':
+#locate index
+
+try:
+index_of_letter = queue[0].index(b)
+
+except ValueError:
+index_of_letter = None
+else:
+c1 = queue[1][index_of_letter]
+
+#calculate time diff 
+tdelta = datetime.strptime(c, FMT) - datetime.strptime(c1, FMT)
+
+df1 = pd.DataFrame({'key' :[b],
+'time' :[c1],
+'key_press_time' :[tdelta.microseconds]})
+data1 = data1.append(df1,ignore_index=True)
+
+#remove key from queue
+del queue[0][index_of_letter]
+del queue[1][index_of_letter]
+if loop>5000:
+break
+
+for i in range(len(data1.index)):
+if i == 0:
+[a1,b1,c1] = data1.iloc[i]
+df1 = pd.DataFrame({'key' :[a1],
+ 'press_time' :[0],
+ 'latency' :[b1]})
+data2 = data2.append(df1,ignore_index=True)
+else:
+[a1,b1,c1] = data1.iloc[i]
+[a2,b2,c2] = data1.iloc[i-1]
+tdelta = datetime.strptime(c1, FMT) - datetime.strptime(c2, FMT)
+df1 = pd.DataFrame({'key' :[a1],
+ 'press_time' :[tdelta.microseconds],
+ 'latency' :[b1]})
+data2 = data2.append(df1,ignore_index=True)
+return data2
+
+#####################################################
+
+
+
+
+'''To remove outliers from the dataset '''
+##############################################
+
+def remove_outlier(df_in, col_name):
+ q1 = df_in[col_name].quantile(0.25)
+ q3 = df_in[col_name].quantile(0.75)
+ iqr = q3-q1 #Interquartile range
+ fence_low = q1-1.5*iqr
+ fence_high = q3+1.5*iqr
+ df_out = df_in.loc[(df_in[col_name] > fence_low) & (df_in[col_name] < fence_high)]
+ return df_out
+
+###############################################
+
+
+
+
+
+
+'''To remove outliers from the dataset '''
+##############################################
+
+def get_data(n_classes,emotion='Happy',neutral=False,cont=False):
+data_list1=[]
+data_list2=[]
+label_list=[]
+data = pd.DataFrame({'key' : [],
+ 'press_time' : [],
+ 'latency' : []})
+data1 = pd.DataFrame({'key' : [],
+ 'press_time' : [],
+ 'latency' : []})
+for i in range(n_classes):
+#Selecting the type of data
+if neutral:
+fldr_name='Neutral'
+elif cont:
+fldr_name='Continuous'
+else:
+fldr_name='Emotional/'+emotion
+#Extracting raw data from the files 
+data1=extract_data(user_details[i][0],user_details[i][1],fldr_name,cont)
+data2 = data1.loc[(data1['press_time'] != 0)]
+#Removes the outliers from the data	with respect to both the attributes
+data=remove_outlier(remove_outlier(data2,'press_time'),'latency')
+for j in range(data.shape[0]):
+#Converted the data to ms
+data_list1.append(data.iloc[j,1]/1000)
+data_list2.append(data.iloc[j,2]/1000)
+label_list.append(i)
+#To store the data in a numpy array in a proper way
+X_data=np.zeros((len(data_list1)-1,2))
+Y_data=np.zeros((len(data_list1)-1))
+X_data[:,0]=np.array(data_list1[0:-1])
+X_data[:,1]=np.array(data_list2[0:-1])
+Y_data[:]=np.array(label_list[0:-1])
+Y_data=np.reshape(Y_data,(len(data_list1)-1,1))
+return X_data,Y_data
+
+###############################################
+
+
+
+
+
+
+'''Main function to verify the Feature extraction '''
+#####################################################
+
+
+if __name__ == '__main__':
+n_classes=5
+X_cont,Y_cont = get_data(n_classes,cont=True)
+print X_cont.shape
+print Y_cont.shape