Biometric: EEG brainwaves

Transcription

1 Biometric: EEG brainwaves Jeovane Honório Alves 1 1 Department of Computer Science Federal University of Parana Curitiba December 5, 2016 Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

2 Introduction Electroencephalogram (EEG) Conclusion Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

3 Introduction EEG Conclusion References Introduction I What is a biometric system? I Types of biometrics I Human identification by physiological and/or behavioral features I Spoofing Jeovane Hon orio Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

4 Introduction Biometric system using brainwaves Analysis of the brain activity Unique for each person Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

5 Introduction Electroencephalogram (EEG) Conclusion Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

6 Electroencephalogram (EEG) What is a electroencephalogram (EEG)? Electro-physiological monitoring method Recording of brain activity Typically non-invasive Measurement of voltage fluctuations by electrodes (gel/dry) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

7 Electroencephalogram (EEG) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

8 Electroencephalogram (EEG) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

9 Electroencephalogram (EEG) Advantages of the brainwaves usage for biometric: High variance inter-class and low variance intra-class (short and long term) Confidential (mental task) Difficult to mimic Sensitive to alterations of the mental state (rejection to a mental state caused by aggressors) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

10 Electroencephalogram (EEG) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

11 Electroencephalogram (EEG) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

12 Electroencephalogram (EEG) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

13 Electroencephalogram (EEG) There are many types of waves based on the frequency (Hz) of the signal Delta: State of deepest meditation and dreamless dream Theta: Most in sleep state or deep meditation Alpha: Resting state of the brain, where the person is in a calm mental state Beta: Normal state for cognitive tasks and attention of the outside world Gamma: Fastest of brainwaves, which is related to simultaneous processing of information in different areas of the brain Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

14 Introduction Electroencephalogram (EEG) Conclusion Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

15 Most biometric works involving EEG follows the procedure below: Extraction of EEG signal Processing of the signal for signal-to-noise improvement Feature extraction Identification/authentication Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

16 Extraction of EEG signal Extraction of these signals depends on: Electrode location Mental state of the person Different equipments and quantity of electrodes were used Commonly, brainwaves in the alpha, beta and/or theta frequencies are the focus Meditation and cognitive activities are realized for extraction of these signals in different frequencies Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

17 Preprocessing of the signal for signal-to-noise improvement This step is not always present An example of preprocessing is the surface Laplacian, for improvement of the cortical activity Frequency filtering Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

18 Feature extraction Overall, the power spectral density (PSD) is used The band, channels and recording durations are variable Modifications on these configurations change considerably the final results Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

19 Identification/authentication In this step, machine learning algorithms like SVM and KNN can be employed Other works stated that this type of techniques lead to time-consuming classifications Techniques like diagonal Gaussian Mixture Models with empirically obtained thresholds were also utilized Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

20 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Portable equipment for recording of signal recording One electrode Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

21 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Total of 40 healthy volunteers (29 male and 11 female) Quiet and clean environment Volunteer in a calm state, on a sofa with closed eyes Two-days recording A diet-analysis was realized with water and coffee Six sessions were realized with intervals of 30 minutes Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

22 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Autoregressive (AR) model prediction coefficients and PSD were calculated for feature vector generation For the AR model, the whole spectrum was used, instead of a frequency band Specifically for the PSD, the frequency below 4Hz and above 33Hz were removed because of noises, after initial experiments Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

23 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Results: The main coffee effect is among 30 minutes to 2 hours after drinking A decrease of almost 10% of the performance was stated, but experiments with different diets and days is necessary Analysis of different recording durations shows that after 3 minutes, a high and stable accuracy was obtained Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

24 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

25 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Results: For classification, the Kohonen s LVQ network, multi-class SVM and knn (k = 1), only and combined with Fisher s linear discriminant analysis (FDA) Hold-out method, dividing 100 times the training and testing data evenly Best results were achieved with knn+fda, with a average accuracy of 97.5% Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

26 Article: EEG-based Personal Identification: from Proof-of-Concept to A Practical System (2010) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

27 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) Consumer-grade EEG sensor Single channel Authentication and identification based on 15 subjects and diverse tasks Analysis of alpha and beta waves Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

28 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

29 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

30 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) A two-day session with repetition of five times was realized for the following tasks (for 10 seconds): Breathing Finger movement Sport Song thinking Eye and Audio tone Colored object counting Pass-thought Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

31 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) PSD extracted Alpha and beta waves segmented from the middle five seconds A 2D power spectrum is compressed to a 1D For each frequency, the median magnitude across time is obtained Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

32 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) Similarity between two signals is computed with the cosine similarity From it, two similarities are computed: self and cross-similarities These similarities are then used for authentication and identification Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

33 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

34 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) For authentication, a three-protocol experiment is employed Baseline protocol Customized Threshold Customized Task Customized Threshold An adapted knn-graph classifier is used for identification, obtaining an accuracy of 22% Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

35 Article: I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves (2013) Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

36 Introduction Electroencephalogram (EEG) Conclusion Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

37 Impostor simulating client tasks Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39 Conclusion A study of EEG biometric was realized From the works analyzed, promising results were obtained Since it is a relative new study, there is few works developed Data sets have few samples Variety and high quantity of tasks are necessary to better reproduce a practical stage Current hardware are not yet practical Experiments with new tasks

38 References 1. Marcel S, R. Millan J. Person Authentication Using Brainwaves (EEG) and Maximum A Posteriori Model Adaptation. IEEE Trans Pattern Anal Mach Intell [Internet] Apr;29(4): Available from: 2. Nakanishi I, Baba S, Miyamoto C. EEG based biometric authentication using new spectral features. In: 2009 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) [Internet]. IEEE; p Available from: 3. Su F, Xia L, Cai A, Wu Y, Ma J. EEG-based Personal Identification: from Proof-of-Concept to A Practical System. In: th International Conference on Pattern Recognition [Internet]. IEEE; p Available from: Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39

39 References 4. Klonovs J, Petersen CK, Olesen H, Hammershoj A. ID Proof on the Go: Development of a Mobile EEG-Based Biometric Authentication System. IEEE Veh Technol Mag [Internet] Mar;8(1):81 9. Available from: 5. Chuang J, Nguyen H, Wang C, Johnson B. I Think, Therefore I Am: Usability and Security of Authentication Using Brainwaves. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) [Internet] p Available from: Mohanchandra K. Using Brain Waves as New Biometric Feature for Authenticating a Computer User in Real-Time. Int J Biometric Bioinforma. 2013;7(1): Jeovane Honório Alves (UFPR) Biometric: EEG brainwaves Curitiba December 5, / 39