audio
audioduration (s)
1.44
8.98
emotion
stringclasses
7 values
label
class label
7 classes
anger
6anger
neutral
4neutral
boredom
3boredom
anger
6anger
boredom
3boredom
sadness
5sadness
boredom
3boredom
anger
6anger
happiness
2happiness
anger
6anger
anger
6anger
neutral
4neutral
sadness
5sadness
boredom
3boredom
boredom
3boredom
anger
6anger
anger
6anger
sadness
5sadness
happiness
2happiness
boredom
3boredom
sadness
5sadness
boredom
3boredom
neutral
4neutral
anger
6anger
disgust
1disgust
happiness
2happiness
happiness
2happiness
disgust
1disgust
happiness
2happiness
neutral
4neutral
neutral
4neutral
boredom
3boredom
boredom
3boredom
anxiety
0anxiety
anger
6anger
anxiety
0anxiety
anger
6anger
boredom
3boredom
disgust
1disgust
happiness
2happiness
boredom
3boredom
anger
6anger
disgust
1disgust
neutral
4neutral
neutral
4neutral
neutral
4neutral
happiness
2happiness
boredom
3boredom
happiness
2happiness
sadness
5sadness
anger
6anger
boredom
3boredom
anger
6anger
anger
6anger
anxiety
0anxiety
anxiety
0anxiety
anxiety
0anxiety
boredom
3boredom
happiness
2happiness
neutral
4neutral
happiness
2happiness
anxiety
0anxiety
happiness
2happiness
sadness
5sadness
happiness
2happiness
neutral
4neutral
disgust
1disgust
anger
6anger
sadness
5sadness
anger
6anger
anger
6anger
anger
6anger
happiness
2happiness
anger
6anger
disgust
1disgust
disgust
1disgust
boredom
3boredom
happiness
2happiness
happiness
2happiness
anxiety
0anxiety
neutral
4neutral
happiness
2happiness
boredom
3boredom
anger
6anger
neutral
4neutral
anger
6anger
sadness
5sadness
happiness
2happiness
anger
6anger
anxiety
0anxiety
neutral
4neutral
neutral
4neutral
happiness
2happiness
neutral
4neutral
sadness
5sadness
disgust
1disgust
neutral
4neutral
happiness
2happiness
disgust
1disgust
anxiety
0anxiety

EmoDB

The EmoDB is the freely available German emotional database, containing a total of 535 utterances. It comprises of seven emotions: 1) anger; 2) boredom; 3) anxiety; 4) happiness; 5) sadness; 6) disgust; and 7) neutral. The data was recorded at a 48-kHz sampling rate and then down-sampled to 16-kHz. We follow the unofficial speaker-independent train/test split from here.

Citations

@inproceedings{burkhardt2005database,
  title={A database of German emotional speech.},
  author={Burkhardt, Felix and Paeschke, Astrid and Rolfes, Miriam and Sendlmeier, Walter F and Weiss, Benjamin and others},
  booktitle={Interspeech},
  volume={5},
  pages={1517--1520},
  year={2005}
}
Downloads last month
44
Edit dataset card