interact : m otion s ensor d riven g esture r ecognition c loud s ervice school of electronic &...

32
INTERACT: MOTION SENSOR DRIVEN GESTURE RECOGNITION CLOUD SERVICE School of Electronic & Computer Engineering Technical University of Crete, Greece Dr. Stelios Sotiriadis K. Stravoskoufos, A. Preventis, E. Petrakis

Upload: roberta-robinson

Post on 23-Dec-2015

214 views

Category:

Documents


0 download

TRANSCRIPT

  • Slide 1
  • INTERACT : M OTION S ENSOR D RIVEN G ESTURE R ECOGNITION C LOUD S ERVICE School of Electronic & Computer Engineering Technical University of Crete, Greece Dr. Stelios Sotiriadis K. Stravoskoufos, A. Preventis, E. Petrakis
  • Slide 2
  • Data! 640K ought to be enough for anybody (1981). Google processes 20 PB = 2.19902326 10 13 kilobytes a day
  • Slide 3
  • Cloud Computing
  • Slide 4
  • The Internet of Things (IoT)
  • Slide 5
  • Slide 6
  • Cloud in Europe: FIWARE Open cloud infrastructure for cost-effective creation and delivery of FI applications and services Develop Cloud Services called Generic Enablers. EU FP7 project. Offers open specification services featuring powerful but user-friendly APIs based on RESTFul standards.
  • Slide 7
  • Generic Enablers (GEs) Speeds up the creation of FI applications. Cloud-based modular components which offer reusable and commonly shared functions. Serve multiple areas of use across various Sectors e.g. health! The fundamental building blocks of FIWARE.
  • Slide 8
  • FI-STAR FP7 FI-STAR Experimentation Sites and Use Case Scenarios. Servicing around 6 Million people throughout Europe. Work in parallel to apply and validate the Core FI- STAR Platform and its subset of GEs. www.fi-star.eu
  • Slide 9
  • XIFI FP7 Crete Node
  • Slide 10
  • XIFI FIWARE Node Crete CPU : 208 Cores RAM: 512 TB HDD: 11.7 TB IaaS services FIWARE Instances Blueprint instances SaaS APIs Feel free to contact me for requesting information!
  • Slide 11
  • Interact! Motion Sensors Cloud Driven Gesture Recognition Cloud open source software based on FIWARE Generic Enablers. It operates FIWARE cloud platform and offer their functionality through RESTful APIs.
  • Slide 12
  • Interact Architecture
  • Slide 13
  • IoT Connectivity & Protocol Adapter Connecting sensor with the FI application components. Uses the Protocol Adapter module to adapt to the specific connectivity protocol that the sensor is using (e.g., Bluetooth).
  • Slide 14
  • Sensor Data Collector Collects the sensor data Converts data into the desirable format (e.g., JSON, XML etc.) and schema so it can be processed in the back-end without causing interoperability problems.
  • Slide 15
  • Connectivity Service Establishes a connection between the front-end and the back-end Transfers the sensor data received from the Sensor Data Collector to the Back-End
  • Slide 16
  • Application Logic Application specific module. Encapsulates the business logic of the FI- application. Handles and processes sensor data by using the Complex Event processing, Cloud storage and Publish subscribe modules
  • Slide 17
  • Complex Event Processing Used for decision-making thorough the analysis of complex conditional events. Processes custom event patterns Based on user defined conditions, decides the flow of the data.
  • Slide 18
  • Cloud Storage Responsible for storing or retrieving sensor data using a database. Functionality should be offered as a RESTful service. Developers should be able to create or retrieve gestures from the database
  • Slide 19
  • Publish Subscribe Responsible for publishing the results of the sensor data processing. Publishes context to context subscribers. Example:
  • Slide 20
  • Identity Management Used for user authentication and access authorization. Users and Developers authorized to access applications and services
  • Slide 21
  • Motion Sensors Sensors that can track the human body or parts of it (skeletal tracking). Asus Xtion Pro MS Kinect Leap Motion
  • Slide 22
  • Interact FI application, based on the FIWARE core platform Example functionality: Gesture Recognition. GESTURE MOTION SENSORINTERACTRESULT Victory"
  • Slide 23
  • Interact Functionalities Recognize Gestures: Static, Motion, Drawing Store Gestures: Developers can use interact to store gestures & create their own gesture pools.
  • Slide 24
  • Interact Functionalities Subscribe to gesture pools: Developers can subscribe to other developers gesture pools in order to use them in their projects. Create gesture pools: Public gesture pool - Can be accessed by everyone - Only super administrators can add new gestures Custom gesture pools - Created by developers
  • Slide 25
  • Interact Functionalities Train the System: Interact features a machine learning mechanism that allows users to train the system to recognize gestures. Implements Naive Bayes Classification Accuracy 75%
  • Slide 26
  • Interact Architecture
  • Slide 27
  • Use Case 1 Gesture to voice: Using Google text to speech API to transform the recognized gestures into voice in order to enable communication with people with speaking disabilities (e.g., deaf-mute).
  • Slide 28
  • Use Case 2
  • Slide 29
  • Use Case 2 Demo
  • Slide 30
  • Use Case 2 Architecture
  • Slide 31
  • Conclusions & Future Work Highlights new means of communication for people with special needs. Interact is online! Ask for more details. Try different machine learning algorithms to increase gesture recognition accuracy.
  • Slide 32
  • THANK YOU! Questions ?