Speaking in public is considered as an essential 21st century skill. Nevertheless, designing effective learning environments aimed to foster presentation skills appears to be a challenge, since acquiring such a skill requires sufficient opportunities for practice and feedback. Previous research demonstrated that practicing and receiving feedback in Virtual Reality (VR) supports students’ learning processes as effective as teacher feedback. However, these VR systems were aimed at developing eye contact and use of voice, while body language is another crucial presentation delivery aspect. Therefore, this research project focuses on constructing an application for the smartphone that supports students’ development in posture and gestures independent of time and place. This application, called Honest Mirror, is a personalized micro-learning environment for practicing public speaking skills in a safe environment. By using AI technology, involving neural networks, data about body language is converted into automatically generated feedback messages supporting students in their presentation development. Moreover, the application aims to meet design criteria regarding: scalability, mobility, effectiveness and adoption in education. During the demo session the application will presented, tested and principles for integrating this app in education will be discussed in the light of the global teacher shortage. Considering the need to educate 69 million new teachers globally by 2030, UNESCO perceives these AI technologies for developing skills as key in reducing the pressure on teachers while guiding individual learners and developing personalized and just-in-time educational programs.
Android Studio (latest version) Android phone or Emulator (Android Studio provides the emulator)
Neural Network one: Tensorflow Lite
Neural Network two: Self-mode (see branch 1,branch 2 and branch 3
Java
Kotlin
XML
Python
The application shows the UI of the homemenu, Eventually leading the user to the next pages.
The user will get prompted to select a video from their phone gallery.
The application uses the URI of this video to take frames.
The application will start the neural networks after all frames are taken, analysing if there are keypoints and poses in the images.
The application will receive an overview of poses and scores of the images.
The application will show the user the drawn version of the overview once the user clicks on results.
The user can click on the button to open a new window.
The application will then show a detailed view of the feedback and the sent video in which the user can look back on themselves.
The application is currently still in beta, but can be downloaded in Android Studio on a physical or digital device. Furthermore, The releases are regularly updated once new versions have been created
[TBA]