In-Vehicle Corpus and Signal Processing for Driver Behavior (eBook)

Contents 6
Contributing Authors 9
Introduction 12
Improved Vehicle Safety and How Technology Will Get Us There, Hopefully 15
1.1 Introduction 15
1.2 Highway Safety 16
1.3 Drivers 16
1.4 Attention, Perception, and System Interfaces 17
1.5 In-Vehicle System Technologies 19
1.6 Concerns for Driver Distraction 19
1.7 Conclusions 20
References 21
New Concepts on Safe Driver-Assistance Systems 23
2.1 Introduction 23
2.2 Driver-Assistance Systems in the Market 24
2.3 Driver-Adaptive Assistance Systems 27
2.4 Driver Assistance with Cooperation Among Vehciles 30
2.5 Discussion 35
2.6 Conclusion 35
References 36
Real-World Data Collection with ‘‘UYANIK’’ 37
3.1 Introduction 37
3.2 ‘‘Uyanik’’ and Sensors 39
3.3 Tasks and Pains 42
3.4 Signal Samples 44
3.5 Gains are Coming: Part 1 46
3.5.1 Audio-Visual Speech Recognition in Vehicular Noise Using a Multi-classifier Approach by H. Karabalkan and H. Erdogbrevean 47
3.5.2 Graphical Model-Based Facial Feature Point Tracking in a Vehicle Environment by S. Coscedilar 50
3.5.3 3D Head Tracking Using Normal Flow Constraints in a Vehicle Environment by B. Akan 51
3.6 Gains are Coming: Part 2 54
3.6.1 Pedal Engagement Behavior of Drivers by M. Karaca, M. Abbak, and M.G. Uzunbascedil 54
3.6.2 Speaker Verification and Fingerprint Recognition by K. Eritmen, M. Imamogbrevelu, and Ç. Karabat 55
3.7 Conclusions and Future Work 56
References 57
On-Going Data Collection of Driving Behavior Signals 58
4.1 Introduction 58
4.2 Design of Data Collection Vehicle 59
4.2.1 Vehicle 59
4.2.2 Microphones 60
4.2.3 Video Cameras 62
4.2.4 Sensors for Driving Operation Signals 62
4.2.5 Vehicle Status Sensors 62
4.2.6 Vehicle Position Sensors 63
4.2.7 Physiological Sensors 63
4.2.8 Synchronous Recording System 63
4.3 Data Collection 64
4.3.1 Tasks 65
4.3.2 Examples of Driving Data 66
4.4 Conclusion and Future Work 66
References 66
UTDrive: The Smart Vehicle Project 68
5.1 Introduction 68
5.2 Multi-Modal Data Acquisition 70
5.2.1 Audio 70
5.2.2 Video 71
5.2.3 CAN-Bus Information 72
5.2.4 Transducers and Extensive Components 72
5.2.5 Data Acquisition Unit (DAC) 73
5.3 Data Collection Protocol 74
5.4 Driving Signals 75
5.5 Driver Distractions 76
5.6 Driver Behavior Modeling 78
5.7 Transcription Convention 78
5.8 Conclusion and Future Work 79
References 79
Wireless Lan-Based Vehicular Location Information Processing 81
6.1 Introduction 81
6.2 Localization 82
6.2.1 Wireless LAN-Based Localization 82
6.2.2 Proximity Approach 83
6.2.3 Triangulation Approach 83
6.2.4 Scene Analysis Approach 84
6.2.5 Accuracy in Outdoor Configuration 84
6.3 Orientation Estimation 85
6.3.1 Difference of Signal Strength 85
6.3.2 Orientation Estimation in Vehicle 86
6.4 Metropolitan-Scale Localization 88
6.4.1 Feasibility of Metropolitan-Scale Localization 88
6.4.2 Locky.jp 90
6.4.3 Current status 91
6.4.4 Related Projects 91
6.4.5 On-Going and Future Work 92
6.5 Conclusion 93
References 93
Perceptually Optimized Packet Scheduling for Robust Real-Time Intervehicle Video Communications 95
7.1 Introduction 95
7.2 The Inter-Vehicle Video Communications Scenario 97
7.2.1 The H.264 Video Coding Standard 97
7.2.2 Analysis-by-Synthesis Distortion Estimation for Video Packets 97
7.2.3 Multimedia Communications over 802.11 99
7.3 The Perceptually Optimized Packet Scheduling Algorithm 100
7.4 Experimental Setup 101
7.5 Results 102
7.6 Conclusions 107
References 107
Machine Learning Systems for Detecting Driver Drowsiness 109
8.1 Introduction 109
8.2 Methods 111
8.2.1 Driving Task 111
8.2.2 Head Movement Measures 112
8.2.3 Facial Action Classifiers 112
8.3 Results 114
8.3.1 Facial Action Signals 115
8.3.2 Drowsiness Prediction 116
8.3.3 Coupling of Behaviors 119
8.4 Conclusions and Future Work 120
References 122
Extraction of Pedestrian Regions Using Histogram and Locally Estimated Feature Distribution 123
9.1 Introduction 123
9.2 Related Research 125
9.3 Kernel Density Estimator With Bayesian Discriminant Function 126
9.3.1 Region of Interest for Processing 126
9.3.2 Modifying Probability Distribution Function 126
9.3.3 Category Estimation by Histograms 128
9.3.4 Kernel Density Estimation with Proposed Preprocessing 128
9.4 Pre-Experiment With Gaussian Shape Distribution 130
9.5 Shape-Dependent Probability Map Template 130
9.5.1 Experiment of Criterion Performance 133
9.6 Conclusion 135
References 135
EEG Emotion Recognition System 137
10.1 Introduction 137
10.2 Emotional Data Collection 138
10.2.1 Experimental Setup 138
10.2.2 Psychological Experiments 139
10.3 Feature Extraction 140
10.3.1 Selection of Experiment Parameters 141
10.3.2 RVM Model 141
10.4 Experimental Results 142
10.5 Conclusions 146
References 147
Three-Dimensional Ultrasound Imaging in Air for Parking and Pedestrian Protection 148
11.1 Motivation - Why Ultrasound? 148
11.2 Signal Model 149
11.3 Image Generation 151
11.3.1 System Setup 152
11.3.2 Data Processing 152
11.3.3 Beamforming 153
11.3.4 Evaluation Criteria 153
11.4 Experiments and Discussion of Results 154
11.4.1 Rough Surface Structure-Continuous Response 154
11.4.2 Smooth Surface - Specular Response 156
11.5 Conclusions 157
References 158
A New Method for Evaluating Mental Work Load In n-Back Tasks 159
12.1 Introduction 160
12.2 Model of Eye Movement 161
12.2.1 Model of VOR 161
12.2.2 Model Identification 162
12.2.3 Identification Results 163
12.3 Method Of Experiment 164
12.3.1 Experiment Procedure 164
12.3.2 Results of Experiment 166
12.4 Conclusion 168
References 168
Estimation of Acoustic Microphone Vocal Tract Parameters from Throat Microphone Recordings 170
13.1 Introduction 170
13.2 Acoustic-Throat Correlation Model 172
13.2.1 Vector Quantization-Based Estimator 173
13.2.2 Hidden Markov Model-Based Estimator 174
13.3 Experimental Results 175
13.4 Conclusions 177
References 177
Cross-Probability Model Based on Gmm for Feature Vector Normalization 179
14.1 Introduction 179
14.2 Memlin Overview 181
14.2.1 MEMLIN Approximations 181
14.2.2 MEMLIN Enhancement 183
14.3 Cross-Probability Model Performance 183
14.4 Cross-Probability Model Based on GMM 185
14.4.1 The E Step 186
14.4.2 The M Step 187
14.5 Normalized Space Acoustic Models 188
14.6 Discussion of Results 188
14.6.1 Results with SpeechDat Car Database 188
14.6.2 Results with Aurora2 Database 191
14.7 Conclusions 192
References 192
Robust Feature Combination for Speech Recognition Using Linear Microphone Array in a Car 194
15.1 Introduction 194
15.2 MFCC Average and Variance Re-scaling 195
15.3 GMM-Based Variance Normalization 197
15.4 Hypothesis-Based Feature Combination of Multiple Inputs 197
15.5 Experimental Results 199
15.5.1 Database and Setup 199
15.5.2 MFCC Average 199
15.5.3 Hypothesis-Based Feature Combination 201
15.6 Conclusions 202
References 203
Prediction of Driving Actions from Driving Signals 204
16.1 Introduction 204
16.2 Driving Signals 205
16.2.1 Types of Driving Signals 205
16.2.2 Database 206
16.3 Predicting Driving Actions 206
16.3.1 Kinds of Driving Actions 206
16.3.2 Methodology of Driving Action Prediction 207
16.4 Experimental Data 208
16.4.1 Driving Action Labels 208
16.4.2 Training Data and Test Data 209
16.5 Results of Prediction Experiments 210
16.5.1 Prediction Performance for Different Driving Signal Input Durations (Experiment 1) 210
16.5.2 Prediction Performance for Different Amounts of Training Data (Experiment 2) 211
16.5.3 Prediction Performance that Considers Individuality of Driving 212
16.5.4 Prediction Performance Using Only One Signal out of five Driving Signals (Experiment 4) 213
16.5.5 Prediction Performance Using Three Useful Signals out of Five 215
16.5.6 Prediction Performance Using Detailed Classification of Driving Actions (Experiment 6) 215
16.6 Conclusion 217
References 217
Design of Audio-Visual Interface for Aiding Driver’s Voice Commands in Automotive Environment 218
17.1 Introduction 218
17.2 Visual Feature Extraction 219
17.3 SNR-Dependent Audio-Visual Information Combination 222
17.3.1 Acoustic Feature Extraction and Estimation of SNR 222
17.3.2 Audio-Visual Model Combination 223
17.4 Experiments and Results 224
17.5 Conclusion 225
References 225
Estimation of High-Variance Vehicular Noise 227
18.1 Introduction 227
18.2 Background 229
18.2.1 Statistical Noise Model 229
18.2.2 MMSE A Priori Noise Estimation 230
18.2.3 Noise Estimation with Speech Presence Uncertainty 230
18.2.4 MMSE A Posteriori Noise Estimation 231
18.2.5 Autoregressive Noise Adaptation 232
18.3 Estimation of High-Variance Noise 232
18.3.1 Speech Presence Probability with Low SNR 232
18.3.2 Proposed Noise Estimation Method 234
18.4 Experiments 234
18.5 Conclusion 237
References 237
Feature Compensation Employing Model Combination for Robust In-Vehicle Speech Recognition 239
19.1 Introduction 240
19.2 CU-Move Corpus 241
19.3 PCGMM-Based Feature Compensation 241
19.4 PCGMM-Based Method Employing Multiple Environmental Models 243
19.5 Noise Transition Model 244
19.6 Experimental Results 245
19.7 Conclusion 248
References 249
Index 250