for SelfDefense Education Manan Mehta Bhushan Muthiyan Hyeran Jeon Kaikai Liu Younghee Park Jerry Gao Gong Chen Jim Kao Department of Computer Engineering ID: 641064
Download Presentation The PPT/PDF document "Intelligent Learning Systems Design" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Intelligent Learning Systems Design for Self-Defense Education
Manan Mehta*, Bhushan Muthiyan*,Hyeran Jeon*, Kaikai Liu*, Younghee Park*, Jerry Gao*, Gong Chen‡, Jim Kao‡*Department of Computer Engineering‡ Department of KinesiologySan Jose State University
1Slide2
Increasing Security Concerns
Increasing unarmed or armed attacksEven college campuses are targets for mass shootingsRun, Hide, and Fight by the Department of Homeland SecurityIndividuals should know how to protect themselvesSelf-defense courses in collegesMartial arts studios2Slide3
Lab with 360 degree camera doesn’t fit well with self-defense education Can capture movement of only single student at a time
Expensive; can’t use the facility outside the labHidden area problem; depending on camera angleSlow; compute-intensive image recognition for the movement modelingLimitations of Existing Solutions3Slide4
Related works in Vision based approach
Inferring the gestures through image segmentation of hand or body jointsICCV13: Real-time articulated hand pose estimation using semi-supervised transductive regression forests CVPR14: Latent regression forest: Structured estimation of 3d articulated hand postureACM Trans on Graphics 2014: Real-time continuous pose recovery of human hands using convolutional networksACM Trans (TIST) 2014: A Real-Time Hand Posture Recognition System Using Deep Neural Networks4Slide5
Related works in Motion based approach
Wearable products: Jawbone UP3 and UP4, FitBit, Apple Watch, Microsoft Band, and GoQii.TuringSense: tennis playersPears: great workouts by the family of coaches mThrow by Motus measures up to 100 metrics points of stress on joints for baseball playersSensoria Smart Socks woven three soft pressure sensors and a magnetic Bluetooth electronic anklet snapsMoov now straps on your wrist to help you analyze running, swimming, cycling and sleeping by providing a unified App.
5Slide6
Related works in Motion based approach
IMU based approachIEEE Sensors 12: MEMS accelerometer based nonspecific-user hand gesture recognitionIEEE Sensors 14: 2D human gesture tracking and recognition by the fusion of MEMS inertial and vision sensors ACM MobiSys 14: Risq: Recognizing smoking gestures with inertial sensors on a wristband6Slide7
Proposed Architecture
7Slide8
Wearable System Design
8Slide9
Major Components
Wearable Device: The designed device contains multiple distributed pea-sized 9-axis inertial measurement units (IMUs) to capture the full motion of forearm, elbow and shoulder. Smartphone App: The smartphone App collects multi-sensor data, eliminates outliers, and feeds the pre-processed sensor data to the estimation engine, where the coordinated movement of the body’s segments are derived. The derived movement data will be fed into the machine learning pipeline to accurately analyze learning effectiveness and sessions in real time, and deliver real-time metrics as the feedback for the trainers. Cloud Dashboard: The Dashboard in the web shows summaries of the training data per-session. The collected training data will be compared with the demo model for the effectiveness and performance evaluation. The dashboard can automatically build the evaluation rules and models just by learning the sensed data from the coach’s demo – combines good and common mistakes- without bringing in IT professionals to write computer code and hand pick gesture features.
9Slide10
Preliminary Results
10Slide11
Thank you
11