This compressed archive contains expressive audio-visual speech (AVTTS) videos in the Greek language. 

The folder called emotion_independent has four subfolders (titled: neutral, angry, happy, sad), with each one containing AVTTS videos of the respective emotion of the title. These videos were generated by training an Hidden Markov Model (HMM) based AVTTS system in a corpus containing utterances of only the corresponding emotion.

The folder called adaptation contains videos of an HMM-based AVTTS system trained on a neutral corpus that has been adapted to another emotion using a number of utterances. The folder has three subfolders (titled: angry, happy, sad) which denote the target emotion each time, and each subfolder has 3 subfolders (titled: 10, 50, 100) which denote the number of sentences used for HMM adaptation.

The folder called interpolation contains videos where two HMM-based AVTTS systems have been interpolated. Each subfolder has a name of the form

weight1_emotion1_weight2_emotion2

where weight1 denotes the interpolation weight of emotion1 and weight2 the interpolation weight of emotion2.

For more information and/or questions please contact Mr. Filntisis Panagiotis Paraskevas at filby@central.ntua.gr.
