1. 2 Abstract - Two experimental paradigms : - EEG-based system that is able to detect high mental workload in drivers operating under real traffic condition.

Slides:



Advertisements
Similar presentations
Electroencephalogram (EEG) and Event Related Potentials (ERP) Lucy J. Troup 28 th January 2008 CSU Symposium on Imaging.
Advertisements

IntroductionMethods Participants  7 adults with severe motor impairment.  9 adults with no motor impairment.  Each participant was asked to utilize.
Brain-computer interfaces: classifying imaginary movements and effects of tDCS Iulia Comşa MRes Computational Neuroscience and Cognitive Robotics Supervisors:
Recognition memory amongst individuals varying in the personality dimensions of Reward Seeking and Impulsivity Chase Kluemper 1, Chelsea Black 1, Yang.
Visualization of dynamic power and synchrony changes in high density EEG A. Alba 1, T. Harmony2, J.L. Marroquín 2, E. Arce 1 1 Facultad de Ciencias, UASLP.
Computer Vision for Human-Computer InteractionResearch Group, Universität Karlsruhe (TH) cv:hci Dr. Edgar Seemann 1 Computer Vision: Histograms of Oriented.
Real-Time Human Pose Recognition in Parts from Single Depth Images Presented by: Mohammad A. Gowayyed.
Pre-processing for EEG and MEG
Artifact (artefact) reduction in EEG – and a bit of ERP basics CNC, 19 November 2014 Jakob Heinzle Translational Neuromodeling Unit.
AdaBoost & Its Applications
Face detection Many slides adapted from P. Viola.
Haptic Signals for Communication under Workload In a primarily visual task, haptic signals can be more resistant to large cognitive workloads than visual.
Real-time Embedded Face Recognition for Smart Home Fei Zuo, Student Member, IEEE, Peter H. N. de With, Senior Member, IEEE.
1 Learning to Detect Objects in Images via a Sparse, Part-Based Representation S. Agarwal, A. Awan and D. Roth IEEE Transactions on Pattern Analysis and.
Robust Real-time Object Detection by Paul Viola and Michael Jones ICCV 2001 Workshop on Statistical and Computation Theories of Vision Presentation by.
The auditory cortex mediates the perceptual effects of acoustic temporal expectation Santiago Jaramillo & Anthony M Zador Cold Spring Harbor Laboratory,
Ensemble Tracking Shai Avidan IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE February 2007.
Face detection and recognition Many slides adapted from K. Grauman and D. Lowe.
1 Integration of Background Modeling and Object Tracking Yu-Ting Chen, Chu-Song Chen, Yi-Ping Hung IEEE ICME, 2006.
Jacinto C. Nascimento, Member, IEEE, and Jorge S. Marques
Rapid Serial Visual Presentation (RSVP) Task (abbreviated sequence) Simulates saccadic vision Used to gauge speed of visual object recognition Thorpe et.
Distinguishing Evidence Accumulation from Response Bias in Categorical Decision-Making Vincent P. Ferrera 1,2, Jack Grinband 1,2, Quan Xiao 1,2, Joy Hirsch.
Face Recognition Using Neural Networks Presented By: Hadis Mohseni Leila Taghavi Atefeh Mirsafian.
EE392J Final Project, March 20, Multiple Camera Object Tracking Helmy Eltoukhy and Khaled Salama.
1 REAL-TIME IMAGE PROCESSING APPROACH TO MEASURE TRAFFIC QUEUE PARAMETERS. M. Fathy and M.Y. Siyal Conference 1995: Image Processing And Its Applications.
EE513 Audio Signals and Systems Statistical Pattern Classification Kevin D. Donohue Electrical and Computer Engineering University of Kentucky.
Convolutional Neural Networks for Image Processing with Applications in Mobile Robotics By, Sruthi Moola.
Theta-Coupled Periodic Replay in Working Memory Lluís Fuentemilla, Will D Penny, Nathan Cashdollar, Nico Bunzeck, Emrah Düzel Current Biology, 2010,20(7):
Change blindness and time to consciousness Professor: Liu Student: Ruby.
Adaptive, behaviorally gated, persistent encoding of task-relevant auditory information in ferret frontal cortex.
Cognitive demands of hands-free- phone conversation while driving Professor : Liu Student: Ruby.
1. Introduction Motion Segmentation The Affine Motion Model Contour Extraction & Shape Estimation Recursive Shape Estimation & Motion Estimation Occlusion.
Participants: 57 children (6-8 years old, 35 boys) participated in experiments. All were schoolchildren in first class of elementary school in Novosibirsk,
1 Methods for detection of hidden changes in the EEG H. Hinrikus*, M.Bachmann*, J.Kalda**, M.Säkki**, J.Lass*, R.Tomson* *Biomedical Engineering Center.
Graz-Brain-Computer Interface: State of Research By Hyun Sang Suh.
Chapter 21 R(x) Algorithm a) Anomaly Detection b) Matched Filter.
Basics of Neural Networks Neural Network Topologies.
Functional Brain Signal Processing: EEG & fMRI Lesson 4
COSC 3461: Module 9 A Principle of UI Design (revisited)
Analysis of Movement Related EEG Signal by Time Dependent Fractal Dimension and Neural Network for Brain Computer Interface NI NI SOE (D3) Fractal and.
Using Feed Forward NN for EEG Signal Classification Amin Fazel April 2006 Department of Computer Science and Electrical Engineering University of Missouri.
Database Comparisons: Age effects. EEG age effects by hemisphere.
The Viola/Jones Face Detector A “paradigmatic” method for real-time object detection Training is slow, but detection is very fast Key ideas Integral images.
Face Image-Based Gender Recognition Using Complex-Valued Neural Network Instructor :Dr. Dong-Chul Kim Indrani Gorripati.
GENDER AND AGE RECOGNITION FOR VIDEO ANALYTICS SOLUTION PRESENTED BY: SUBHASH REDDY JOLAPURAM.
RCC-Mean Subtraction Robust Feature and Compare Various Feature based Methods for Robust Speech Recognition in presence of Telephone Noise Amin Fazel Sharif.
Principal components analysis (PCA) as a tool for identifying EEG frequency bands: I. Methodological considerations and preliminary findings Jürgen Kayser,
EEG processing based on IFAST system and Artificial Neural Networks for early detection of Alzheimer’s disease.
FMRI and Behavioral Studies of Human Face Perception Ronnie Bryan Vision Lab
The role of visuo-spatial working memory in attention to eye gaze Anna S. Law, Liverpool John Moores University Stephen R. H. Langton, University of Stirling.
A direct comparison of Geodesic Sensor Net (128-channel) and conventional (30-channel) ERPs in tonal and phonetic oddball tasks Jürgen Kayser, Craig E.
Maxlab proprietary information – 5/4/09 – Maximilian Riesenhuber CT2WS at Georgetown: Letting your brain be all that.
Evaluation of Gender Classification Methods with Automatically Detected and Aligned Faces Speaker: Po-Kai Shen Advisor: Tsai-Rong Chang Date: 2010/6/14.
Event-Related Potentials Chap2. Ten Simple Rules for Designing ERP Experiments (2/2) 임원진
CONTENTS:  Introduction.  Face recognition task.  Image preprocessing.  Template Extraction and Normalization.  Template Correlation with image database.
Methods for Dummies M/EEG Analysis: Contrasts, Inferences and Source Localisation Diana Omigie Stjepana Kovac.
Melanie Boysen & Gwendolyn Walton
[Ran Manor and Amir B.Geva] Yehu Sapir Outlines Review
Bag-of-Visual-Words Based Feature Extraction
Verifiability and Action verb Processing: An ERP Investigation
Automatic Sleep Stage Classification using a Neural Network Algorithm
Volume 69, Issue 3, Pages (February 2011)
Perceptual Echoes at 10 Hz in the Human Brain
EE513 Audio Signals and Systems
Visual Processing in Fingerprint Experts and Novices
Dynamic Causal Modelling for M/EEG
Machine Learning for Visual Scene Classification with EEG Data
Encoding of Stimulus Probability in Macaque Inferior Temporal Cortex
Auditory Morphing Weyni Clacken
Random Neural Network Texture Model
Presentation transcript:

1

2 Abstract - Two experimental paradigms : - EEG-based system that is able to detect high mental workload in drivers operating under real traffic condition ① An auditory workload scheme ② A mental calculation task - performance is strongly subject-dependent; however, the results are good to excellent for the majority of subjects

3 Introduction - Important issue in fields : pilots, flight controllers, operators of industrial plants - Be integrated with existing systems to maximize the performance - Approach #1 : A closed-loop system The system’s interaction with the operator is adjusted According to the operator’s mental workload - Approach #2 : The workload detector as an objective measure of mental workload To develop improved modes of human-machine interaction - Driving a car while performing additional tasks that model interaction with the car’s systems To obtain a system that is able to measure and mitigate mental workload (1) in real time and (2) in a real operational environment

4 The Experimental Setup - A system that is able to measure and mitigate mental workload in real time and in real operational environments. ⊙ the execution of secondary tasks not related to driving ⊙ Twelve male and five female subjects age 20 to 32 years old ⊙ Approximately 100 km/h on the highway in moderate traffic conditions ⊙ Not to speak during the experiment in order to avoid additional workload - Perform three types of tasks : #1 Driving the vehicle #2 An auditory reaction time task #3 A High mental workload task

5 #3 - 2 An auditory reaction time task #3 - 1 A mental calculation task ① German words links (left) and rechts (right) were randomly presented every 7.5 s ② Asked to count down in steps of twenty-seven ③ After two minutes, the subjects were asked for the final result ① A female news reader and a male voice reciting from a book ② The subject were instructed to follow the latter. ( To verify whether the subjects were engaged or not, they had to answer related questions) Subject had to their attention to one of two simultaneously presented voice recordings, replicating a situation in which several vehicle occupants are talking at the same time. #2 An auditory reaction time task ① Give three digit random number (between 800 and 999). ② Press corresponding buttons

6 A crucial purpose of the experiment is to investigate whether the output of the workload detector can be used to control the secondary task

7 Online Detection of Mental Workload #1 The Workload detector

8 ⊙ Artifact channel removal Such broad-band differences are characteristic for muscle artifacts (> 20 Hz) or eye artifacts (< 6 Hz). The channels that exhibit those differences are excluded. ⊙ Selection subset channel Subset is one of four candidate sets that potentially include frontal, occipital, and temporal scalp positions. ⊙ Spatial filtering Each of the selected channels is normalized by the common median reference signal (the median of all channels is subtracted from each channel) ⊙ Bandpass filtering bandpass filter using one of the bands listed above

9 ⊙ Classification Linear model whose parameters are computed by standard linear discriminant analysis (LDA) of the feature vectors The output is scalar value : Low workload (values below zero) High workload (values above zero) high and low workload, by means of a threshold scheme that employs a hysteresis, which makes the classification substantially more robust (m l < m h )

10 ⊙ Parameter Calibration To find suitable values for all the previously mentioned subject- and task-specific parameters, they use the well known cross-validation Technique. This procedure is performed for each possible combination of parameter candidates in the feature extraction part (EEG channel subset, spatial filter, frequency band, window length)

11 Result #1 Neurophysiological Interpretation

12 #2 Accuracy of the Workload Detector Auditory workload : (95.6% correct) Mental calculation workload : (91.8% correct)

13 #3 Performance Improvement

14

15 Abstract Methodology for relating neural variability to response variability, describing studies for response accuracy and response latency during visual target detection. Introduction - Analysis of trial-averaged ERP in EEG has enabled one to assess the speed of visual recognition and discrimination in terms of the timing of the underlying neural processes. - More recent work has used single-trial analysis of EEG to characterize the neural activity directly correlated with behavioral variability during tasks involving rapid visual discrimination. - These results suggest that components extracted from the EEG can capture the neural correlates of the visual recognition and decision making processing on a trial-by-trial basis. ‘ Cortically-coupled computer vision’

16 Linear methods for single-trial analysis - The goal of a BCI system is to detect neuronal activity associated with cognitive events. - Identify only one type of event (visual target recognition, differentiate this from other visual processing) - 64channels, 1000Hz, time window is 0.5s (32000 samples, dimensional feature vector) Sampling1000Hz (up to 500Hz) => L=50 time window, 20Hz (up to10Hz), 640 dimension - The task is a binary classification - The EEG activity is recorded as D x T values : D is the number of channels T is the number of samples - For reasonable classification : (1) Reduce the trial-to-trial variability by filtering the signal to remove 60Hz interference and slow drifts. (2) Reduce the dimentionality of the problem by stepping classification window every L-th sample. L = 50, the signal of interest is at 10Hz while faster signal variation are considered noise. 1, 2, ,499,500 D T 1, 2, ,9,10 D T

17 X is projected onto a single dimension y. (X is a matrix of channels by samples, and y is a row vector containing multiple samples). In this equation the inner product computes the average over trials and samples. When the intensity y averaged within the specified time window is used as classification criteria we achieve on this data an Az-value of 0.84.

18 EEG correlates of perceptual decision making - Using single-trial linear discrimination analysis To identify the cortical correlates of decision making during rapid discrimination of images - A series of target (faces) and non-target (cars) trials are presented in rapid succession - Stimulus evidence is varied by manipulating the phase coherence of the images - Using a set of 12 face (Max Planck Institute face database) and 12 car grayscale images (image size 512 x 512 pixels, 8-bits/pixel).

19 Early component ( = 170 ms following stimulus) Late component ( > 300 ms following stimulus)

20 Identifying cortical processes leading to response time variability - Studying visual target detection using an RSVP paradigm - During this task, participants are presented with a continuous sequence of natural scenes. RSVP : rapid serial visual presentation - 4 blocks of 50 sequences - Each sequence has a 50% chance of containing one target image with one or more people in a natural scene - These target images can only appear within the middle 30 images of each 50 image sequence. - The Each image was presented for 100 ms

21 - Gaussian profile : - Linear regression coefficients : latency of the component activity