Volume 119 No. 12 2018, 16383-16389 ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu ijpam.eu Driver Fatigue Detection Using Neurosky Mindwave Headset N.R. Raajan 1, Isra.A.R, Greeta S 1, Hemapriya.N 1, Sujanth Narayan K.G 1, Madhupriya.G 1,K.Hariharan 2 1 School of Electrical and Electronics Engineering, SASTRA Deemed University, Tamil Nadu, India 2 School of Computing, SASTRA Deemed University, Tamil Nadu, India nrraajan@gmail.com,isra12110@gmail.com, greetagreet@gmail.com, hemapriyakuppan@gmail.com, kgsn96@gmail.com, madhupriya019@gmail.com Abstract- Major road accidents occur, when the driver falls asleep and drive the vehicle. Sleep state of human being is divided into five stages. The major cause of accidents is the fatigue state of driver. Signal from brain is claimed to be the EEG signal. It reflects the fatigue state. Neurosky Mindwave headset is used to capture the brainwaves. It is portable device and is in turn connected to a PC or mobile which helps in monitoring the process. This proposal involves the detection of fatigue state and to analyse the brainwaves. The interference is reduced by the band stop, Hilbert-Huang transform and the median filter. Percentage power spectral density yields the eigen value. FFT is also used. Experiment results yield the signal strength, values of brainwaves, FFT results and the raw signal. Keywords-Fatigue; EEG; Thinkgear; Interference; brainwaves; drowsiness 16383
I INTRODUCTION Brain is made up of tiny cells called the neurons. The neurons are responsible for the entire cap acity of humans. When they communicate with each other there arises the neuro impulse which is claimed to be the EEG signal. The Neurosky headset is placed on the scalp region. They mainly work because of the sensors in it. The EEG headset contain a sensor that should be placed over the forehead region. The region is claimed as FP1.The technique of capturing the EEG signal falls under two categories. Former one is the Invasive and the other one is Non-Invasive. Invasive technique involves tiny surgery over the scalp and it is painful. The electrodes are placed over the chosen areas and the capturing is done. The wet sensors are used here. Under few old techniques the body, eye and head movements were considered. The proposed method falls under the Non -Invasive technique. The sensors of Neurosky Mindwave Headset are dry in nature. The sensor 1 is placed over the forehead region (FP1). Human brain has the well expanded frontal pole area which are responsible for the cognitive abilities like memory, visual processing, reasoning, focusing and attention. The frontal pole involves the lateral frontopolar areas and the medial frontopolar areas. Sleep states involves four stages followed by the Non -Rapid Eye Movement stage. There is phase of sleep called the microsleep where the person completely becomes unconscious. This phase is entirely responsible for the road accidents when the fatigued person is involved. Butterworth filter and the median filter are used to remove interference and artifacts. Artifacts are the signal components other than the EEG signal. Artifacts are classified into two types based on the origin. The signals arising from lungs, heart and other organs are called the artifacts. It may be due to the external sources present near the driver. The brain waves include delta waves, theta waves, alpha waves, gamma waves and beta waves components. The delta waves originate from 0 to 4 HZ and (100-200) µv. It is observed during deep sleep, unconscious state. The theta waves range from 4 to 7 HZ and (30µv). It is observed during emotional pressure and interruptions. Alpha waves range from 8 to 13 HZ and (30-50) µv. It is observed while resting and calm state. Gamma waves range from 14 to 30 HZ and (5-20) µv [1]. It is observed whiling paying selective attention. Beta waves start from 14-30 HZ and is observed whiling thinking and at conscious state. BP neural network involves forward propagation of the information and anti-propagation of the error [2]. The input received by the neuron reaches the middle information processing layer and finally to the last layer. Thus, the Forward propagation is done. Anti- Propagation is also observed at the output layers. II METHODOLOGY The Neurosky Mindwave headset can capture the raw brainwave which includes alpha, beta, theta, gamma and delta. The Thinkgear Connector is the application that interfaces the PC and the headset. It uses RF to transfer the data packets. Serial communication protocol is used. Thinkgear library is also involved. Mindwave headset baud rate is 57. The headset is built in with ear clip. It serves for the grounding. Low pass filter has the cut off value of 40Hz. Notch filter eliminates 60 Hz noise in the signal [3]. The obtained signals can be plotted using the software called Neuroview or the Datplot software. EEG power spectrum can be calculated using the FFT. Increasing the no of EEG channels helps in analysing the energy shifts on various bands. Delta and theta waves are stable under fatigue state where alpha waves decrease marginally. The length of the median filter is 100.It is a nonlinear filter that degrade the interference pulses and maintains the original properties of the signal. It is used under the Preprocessing stage. The output of the filter can be given as ) =, n= 2k+1 (odd) -----(1) +) / 2, n= 2k ( even) Hilbert Huang Transform can decompose a nonlinear signal component into small Intrinsic Mode Function components (IMF) and the instantaneous data is obtained by the EMD process. The minimal and 16384
maximal values of x(t) of the EEG signal is found out and the m(t) envelope curve is also taken. The mean value of the envelope curve is calculated. x(t)-m(t) -------(2) The value of when it is found inline with the IMF then it is taken as the first order IMF component. Now the residual can be calculated. It is obtained by subtracting the x(t) which is the original signal and the first order IMF component. The original signal is the summation of all the IMF and the final residual value. The instantaneous phase and the instantaneous frequency can be written as, =((h(t)/x(t))----(3) f(t) = 1/2 A/D (d/d(t))-----(4) conversion The EEG signal is taken from 3 to 30 Hz and each IMF component obtained is filtered based on the condition set. If the component value is lesser or greater than the threshold value, then it is neglected. Percentage Power Spectral Density is calculated after the FFT. PSD = FFT x(t) 2 -------(5) After the HHT process, the new signal components (t) and the EEG signal is the linear combination of all the new components. The signal intensity can vary due to the age, gender. The FFT is the easier way to compute the DFT that reduces the number of computations needed for N points. Fig: 1 Plot of the raw EEG signal Extracted EEG signal from Human Data Acquisition Drowsy state detection Spectrum Analysis Fig 2: Block diagram of the proposed diagram The relationship between the EEG and the fatigue state can be analysed by comparing the PVT and the standard sleep data table with the obtained database. The parameter TD is also taken for the reference. The value ranges from 0 to 1 and each value suggests a different state of the driver [6]. 16385
III RESULTS Fig :3 Raw wave Value Analysis Fig:4 Brainwaves Ranging Fig:5 Processed EEG Signal Fig:6 FFT Power Spectrum Plot 16386
COMPARISON OF THREE DIFFERENT TESTS Table No: 1 Alpha Waves Range LOW ALPHA HIGH RAW WAVE ALPHA 1.07802E-41 4.95639E-42 200 2.51771E-41 2.1992E-41 48 3.6742E-41 4.40386E-41-172 Table No: 1 Alpha Waves Range LOW BETA HIGH BETA RAW WAVE 2.40897E-41 5.65284E-42 200 9.62692E-42 8.56614E-42 48 4.40022E-41 1.31582E-41-172 Table 3: Delta and Theta Waves range DELTA THETA RAW WAVE 4.17824E-40 7.16414E-41 200 1.86478E-40 3.97899E-41 48 1.9094E-39 9.98327E-41-172 Table 4: Gamma Waves Range LOW HIGH RAW WAVE GAMMA GAMMA 4.04975E-42 1.78946E-42 200 7.54879E-42 1.51046E-41 48 7.93976E-42 3.63497E-42-172 IV CONCLUSION Human safety measures are to be taken in every innovation. While driving, the fatigue state based on the body characteristics and the outer look is of low accuracy. Based on the EEG signal and the handy Neurosky Headset, the fatigue state of the driver can be well determined. The signal is converted from time domain to the frequency domain using FFT. Percentage power spectral density can be used to extract characteristics of the EEG. The energy of the various bands is known then. Once the portable headset is wearing by the driver. The monitoring unit can predict the fatigue state using the values of various brainwaves. The signal strength can be analysed. And it has the strong time resolution. The interference level processing is also effective and with goo d accuracy. The final state from the results can be considered to further processing like stopping the vehicle depending upon the traffic and locality or informing the Police or tracking the location of the vehicle. 16387
V REFRENCES [1]Yuan Wang,Yan Zhang, Dan Liu, Driving Fatigue Detection Based on the EEG Signal, School of Electrical Engineering and Automation, Harbin Institute of Technology, Harbin, 2015 (china) [2]Zhizhong Zhao, Haiping Xin, Yaqiong Ren, Xuesong Guo Application and Comparison of BP NeuralNetwork Algorithm in MATLAB, Province-Ministry Joint Key Laboratory of Electromagnetic Field and Electrical Apparatus Reliability, Hebei University of Technology, Tianjin, 300130, 2010 (China) [3] Upasana Sinha, Kamal K. Mehta, PhD,A.K. Shrivastava, PhD Real Time Implementation for Monitoring Drowsiness Condition of a Train Driver using Brain Wave Sensor, Dr.C.V. Raman University Department of Computer Science Engineering Bilaspur, 2016, India. [4] Nandhita, Christy Persya, EEG Based Brain Controlled Robo and Home Appliances,Department of ISE, BNMIT, Bangalore 560085, Karnataka, India 2017 [5] Li Man and Meng Hui-ling, A Method of Driver Fatigue Detection based on Multi-features,Engineering College, Xi'an International University, Xi an, China 2015. [6] Nikita Gurudath, H. Bryan Riley, Drowsy Driving Detection by EEG Analysis Using Wavelet Transform and K-Means Clustering,School of Electrical Engineering and Computer Science, Ohio University, Athens, Ohio 45701, USA 2014. [7] Zhendong Mu, Jianfeng Hu and Jianliang MinDriver, Fatigue Detection System Using EEG signals Based on Combined Entropy Features, Centre of collaboration and Innovation, Jiangxi University of Technology, Nanchang 330098, Jiangxi, China 2017 [8] F.S.C. Clement, Aditya vashistha, Milind E Rane, Driver Fatigue Detection system Electronics Department, VIT Pune 2015 [9] S.Shanmugapriya, 2B.S.Sathishkumar RECOGNITION OF HUMAN IDENTITIES USING ENHANCED KNUCKLE PATTERN FEATURE International Journal of Innovations in Scientific and ISSN: 2347-9728(print) Engineering Research (IJISER) 16388
16389
16390