ADHD indicators modelling based on Dynamic Time Warping from RGB data: A feasibility study Antonio Hernández-Vela, Miguel Reyes, Laura Igual, Josep Moya, Verónica Violant, and Sergio Escalera

ADHD: Attention deficit hyperactivity disorder 2 InattentionHyperactivityImpulsivity

Outline 1.Introduction 2.Methodology 3.Results 4.Conclusion 3

Introduction Video-based behavior analysis for ADHD diagnosis in children between 8-11 years. 4

Introduction Behavior analysis  Human pose information along time 5 Head Body Hands time Gestures Inattention Hyperactivity Impulsivity 1. Data acquisition 2. Feature extraction: Human Pose 3. Gesture detection

Outline 1.Introduction 2.Methodology 1.Data acquisition 2.Feature extraction 3.Gesture detection 3.Results 4.Conclusion 6

Data aqcuisition 7 Microsoft’s Kinect RGB + Depth Invariant to color, texture and lighting conditions Human pose directly obtained

Feature extraction: Human Pose 8 RGB + Depth Body skeleton 42-dimensional vector: 14 joints × 3 spatial dimensions

Gesture detection 9 Dynamic Time Warping (DTW)

Threshold computing 10 Leave-one-out similarity measure between different samples and gestures G1 G11G12…G13 G2 G21G22…G23 … Gn Gn1Gn2Gn3 G1 1 Different gestures Different samples

Outline 1.Introduction 2.Methodology 3.Results 4.Conclusion 11

Results 12

Results 13

Outline 1.Introduction 2.Methodology 3.Results 4.Conclusion 14

Outline 15 1.Introduction 2.Methodology 3.Results 4.Conclusion

Conclusion 16 Methodology for gesture segmentation and recognition at the same time. First results indicate the objectives are feasible. Future work: Automatic callibration Feature weighting (body joints)

ADHD indicators modelling based on Dynamic Time Warping from RGB data: A feasibility study Antonio Hernández-Vela, Miguel Reyes, Laura Igual, Josep Moya, Verónica Violant, and Sergio Escalera Thank You! Questions?