- Open Access
- Authors : Chitra Agnihotri, Jyoti Mengade, Bhumika Sevak, Vaishnavi Utekar
- Paper ID : IJERTCONV9IS03001
- Volume & Issue : NTASU – 2020 (Volume 09 – Issue 03)
- Published (First Online): 13-02-2021
- ISSN (Online) : 2278-0181
- Publisher Name : IJERT
- License: This work is licensed under a Creative Commons Attribution 4.0 International License
Cognitive based Lie Detection
Chitra Agnihotri Computer Department KJ Somaiya, Vidyavihar Thane, India
Jyoti Mengade Computer Department KJ Somaiya, Vidyavihar Thane, India
Bhumika Sevak Computer Department KJ Somaiya, Vidyavihar Thane, India
Vaishnavi Utekar Computer Department KJ Somaiya, Vidyavihar Thane, India
Abstract :- In this paper different cognitive parameters like eye blink rate and analysis of the persons speech frequency to detect the deception along with heart rate using ECG machine to design a lie detection system. A recent study found that lying takes longer than telling the truth, and thus the time to answer a question may be used as a method of lie detection. This paper aims to detect the humans unsatisfactory behaviour to examine to detect the lie. Along with the mentioned parameters this work implements machine learning model which will help to classify the liars into one group and not liars into separate group.
There was a time when people used to lie to simply avoid the consequences of ones actions. But in Todays fast paced world, lies have become a tool to escape from the critical situation, taking the easy way out. Lying hides a persons true image to others. Lies are fine in fiction but not in contemporary world. Once a lie has been told, there can be two alternative consequences: it may be discovered or remain undiscovered. A recent study found that lying takes longer than telling the truth, and thus the time to answer a question may be used as a method of lie detection. To detect the humans unsatisfactory behavior is what we aim to examine.
The idea here is to develop Cognitive Sensing Based Lie Detection which includes parameters like eye blink rate and analysis of the persons speech frequency to detect the deception along with heart rate using ECG machine. The objective of this work is to implement machine learning model which will help to classify the liars into one group and not liars into separate group. However, the results of this analysis cannot be trusted to be completely correct as the person might be clever enough to trick the system by hiding their emotions.
OUTLINE OF THE PAPER :
First of all, we will present the study of various system proposed by the different authors. And after that we will present our proposed system and details of the different hardware component used in the system.
The organization of the rest of the paper is as follows : section 3 presents Literrature survey based on the proposed system in , , , .. Section 4 presents the system proposed in this paper for lie detection. Section 5 presents the detailed
implementation details and working of different component used in the system. And finally, section 6 presents conclusions.
Sabu George et al.  have used face based deception detection system wherein blink rate is used as an indicator of deception as researchers concluded that during deception, blink rate increases. The system here analyzes blink count and blink duration of truth and lie responses of various subjects while answering interview questions. The participants were given 10 set of control questions to answer, wherein facial expressions were recorded via high speed camera (Prosilica GE680) which captures 200 frames per second for AU45 detection which is the action unit for eye blink (facial muscle movements).
K. Meena et al.  have performed guilty knowledge test (GKT) by recording the data, pre-processing the data where basic filtering is applied to reduce the percentage of noise. Here EEG signals are used as an indicator to detect a deception. Also MFCC (Mel Frequency Cepstral Coefficients) is used for feature extraction of all the EEG signals. All the features extracted are given to NN Classifiers for classification of liars into separate classes. Statistical Features like RMS (Root Mean Square), Power, Variance are calculated for results.
Suresh et al.  have used ANN, since it is one of the best methods for lie detection, also with that they have used parameters like blood pressure rate, eye blink rate, lip movements, hand and leg movements and hardware like ECG, EEG which all necessitate to collect without the knowledge of the suspect. All the parameters were used to calculate the results which were presented in the form of polygraph. Fuzzy logic is also used to analyse the persons truth words.
Amanda Chow et al.  have used NLP techniques to determine whether the individual is lying or not, by giving an audio recording of them speaking. They extracted acoustic, lexical, and prosodic features using CSC corpus (which consists of 32 hours of speech data from 32 english speakers) and fed them into several ML Algorithms for deception. They have also used MFCC(Mel Frequency Cepstral Coefficients)
to extract the person's voice feature basically to identify the uniqueness of the persons speech.
Fig .1 Proposed System
Fig. 2 ECG machine connector
Investigator will then suspect the accused person by interviewing with a set of questions called a GKT (Guilty Knowledge Test). Once the GKT is done, processing of the data will be performed where for a given duration of time, the suspects eye blink rate, heart rate and voice frequency under a normal condition will be calculated and stored in the database (MySQL). Then the system will be trained using the Random Forest Classifier. Once the training of the data is done, suspect will be cross questioned and the responses to which will be used as a testing data. Here a threshold is set for the eye blink rate, heart rate, and voice frequency of the current accused/suspected person based on his / her behavior under normal circumstances (training phase).
For testing of a lie, we will feed the classifier with the testing data to compare it with training data, if the value for eye blink rate, heart rate and voice frequency for the current testing data
Fig. 1 shows the system proposed in this paper. The suspect have to sit in front of the laptop where the facial expressions of the suspect will get captured using the laptops web-camera and the answers given by the suspect with respect to the questions asked by the investigator that are in the form of input voice will be recorded using laptops headphone/microphone which will be further extracted using MFCC(Mel Frequency Cepstral Coefficients) for feature extraction and the suspect has to wear the ECG machine which will help the system to analyse the heart rate. These are the parameters which are necessary for the system and will help the system to detect the deception.
The ECG machine requires the 3 leads to be connected as shown in fig. 2. So the pulse signals from the human body will pass through the amplifier which does the filtration of excess noise and will go through the Arduino to the system in the form of frequency.
falls above the threshold, it will be assumed the suspected person is lying otherwise the suspected person is not lying.
Camera – Laptops webcam for capturing eye blinks which results in capturing the eye blink rate whether it is decreasing and stable or increasing to detect lies. Fig. 3 shows how the camera will capture an input to the system. That video will be further processed to examine the eye blinks.
Fig. 3 Input captured by Camera
Mic Input : Laptops Headphone/Microphone which is required by the system for extracting the voice frequency hether the suspects voice is trembling or shuddering.
ECG: The Generic ECG module AD8232 Measurement Pulse heart monitoring sensor Kit for Arduino is used to measure the electrical activity of the heart.
Specification of AD8232:
AD8232 as shown in Fig. 4 adopts double poles high-pass filter to eliminate the motion artifacts and electrode halfcell potential.
The usage of AD8232 is shown in Fig. 5 with the connections that are required to do with the suspect and with the system. Fig. 6 shows the input getting from the AD8232 in the form of waves..
Fig. 4 AD8232
Fig. 5 Connection of AD8232 with suspect and system
Fig. 6 ECG Signals Waveform getting from the ECG module AD8232 connected to suspect.
Noise Remover : AD8232 hardware is used as an operational amplifier to build a three pole low pass filter, eliminating extra noises. Fig. 7 shows the connection of noise remover with the ECG device.
Fig. 7 Noise Remover
Arduino: The ECG Module AD8232 Heart ECG Monitoring Sensor Module Kit for Arduino is designed to extract, amplify, and filter small bio-potential signals in the presence of noisy conditions; such as those created by motion or remote electrode placement. Arduino board is shown in Fig. 8.
Fig. 8 Arduino Board
Specifications of Arduino board:The current version of Arduino Uno comes with USB interface, 6 analog input pins, 14 I/O digital ports that are used to connect with external electronic circuits. Out of 14 I/O ports, 6 pins can be used for PWM output. Details are given as below for attaching ardiuno with ECG machine:
How data is stored: Data is stored inside mysql using pymysql library.
Each time recorded data will be inserted to database.
How data is used: We have recorded heart rate using ECG, eye blinks and voice frequency using MFCC algorithm. Based on this parameter we are detecting whether user is lying or not. If parameters goes beyond threshold value then result will be lie.
Training Algorithm: Random forest algorithm is used for detecting whether a person is lying or not. For detecting purpose we are checking the audio file of answers which is given by the suspect with similar trained data of voice
Training Data person based: If one person training is done it would not go inside algorithm, threshold value will work in this case. We are not considering multiple person based training data.
Results: We haven't reached till the final result yet we are still required to complete the proposed model. However we have completed with taking input and capturing data which is given in the implementation details.
This system is based on building a Lie Detector capable of detecting, monitoring and transmitting suspects physiological parameters such as eye blinking, pitches of voice, heart rate, etc. and to create a system that tells whether a suspect is lying or not based Machine Learning Algorithm.
For the implementation of the proposed system, devices like ECG recorder, Camera and mic are used for recording the data and Database approach is used for storing the data and Machine learning is used to make the system capable to classify / predict whether a person is lying or not.
The main idea or motive behind the project is to tell whether a person is lying or not for which there is no exact way to identify. But with the correct implementation of this device we can expect a rather close to perfect result for most of the cases.
CONFLICT OF INTEREST:
The Author(s) declare that there is no conflict of interest regarding the publication of this manuscript.
Tony Yiu, Understanding Neural Networks, on June 2,2019, [Online] available: https://towardsdatascience.com/understanding- neural-networks-19020b758230
Vikas Gupta, Face detection- OpenCV, dlib and Deep Learning on Oct 22,2018, [Online] available: https://www.learnopencv.com/face- detection-opencv-dlib-and-deep-learning-c-python/
Alexander Mordvintsev & Abid K., Introduction to OpenCV-Python Tutorials, on 2018, [Online] available: https://opencv-python- tutroals.readthedocs.io/en/latest/py_tutorials/py_setup/py_intro/py_i ntro.html
Adrian Rosebrock, Eye Blink detection with OpenCV,Python & dlib on April 2017, [Online] available: https://www.pyimagesearch.com/2017/04/24/eye-blink-detection- opencv-python-dlib/
Sabu George, Manohara Pai M.M ,Radhika M Pai, Samir Kumar Praharaj, Eye Blink Count and Eye Blink Duration Analysis For Detection Deception 2017
K. Meena, K. Veena, Lie Detection System using Input Voice Signal, International Journal of Pure and Applied Mathematics, Volume 117, No. 8 2017, 121-125.
Dr. A. Suresh and H. Shaheen, Speech Stress Analysis Based on Lie Detector for Loyalty Test, International Journal of Printing,
Packaging & Allied Sciences, Vol 4, No. 1, December 2016
Amanda Chow and John Louie Detecting lies via Speech Patterns, Department of Computer Science, Stanford University 2015.
Sharon Leal, Aldert Vrij, Blinking During and After Lying, Journal of Nonverbal Behaviour, Vol. 32, pp. 187-194, 2008.
Kyosuke Fukuda, Eye blinks: new indices for the detection of deception, International Journal of Psychophysiology, Vol.40, pp. 239-245, 2001.