RETTL UCI FIU

Automated robotics training in VR: Concept and Affect Detection.

Confidence Estimation

Requirements

Use Dockerfile to build an docker image and create a new container.

docker build -t confidence .
docker run -itd -v "your_working_dir":/root/home --name confidence confidence:basic

Then start our new container.

docker exec -it confidence /bin/zsh

See special requirements in each subfolder.

Tasks

• Create file name convention.
• Write python code to segment recordings in 5s clips.
• Use google speech-to-text to transcript 5s clips and store transcriptions
• Save the transcription for each 5 seconds clips, with the same naming conventions.
• Extract and store acoustic features of clips (OpenSmile library), with the same naming conventions.
• Concatenate student id, clip id, acoustic features, transcription content to a csv file.

Naming Convention

First directory: studentId

Second directory: studentId-taskId

Third directory: studentId-taskId-clipId

For each clip: studentId-taskId-clipId

Extract Features

Extract and store acoustic features: using compare emotion, EGEmaps configuration files.

Goal: find the features which are most correlated to the labeled confident extend.