• Description:

SAVEE (Surrey Audio-Visual Expressed Emotion) is an emotion recognition dataset. It consists of recordings from 4 male actors in 7 different emotions, 480 British English utterances in total. The sentences were chosen from the standard TIMIT corpus and phonetically-balanced for each emotion. This release contains only the audio stream from the original audio-visual recording. The data is split so that the training set consists of 2 speakers, and both the validation and test set consists of samples from 1 speaker, respectively.

  • Homepage:

  • Source code:

  • Versions:

    • 1.0.0 (default): No release notes.
  • Download size: Unknown size

  • Dataset size: 259.15 MiB

  • Manual download instructions: This dataset requires you to download the source data manually into download_config.manual_dir (defaults to ~/tensorflow_datasets/manual/):
    manual_dir should contain the file This file should be under Data/Zip/ in the dataset folder provided upon registration. You need to register at in order to get the link to download the dataset.

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'test' 120
'train' 240
'validation' 120
  • Features:
    'audio': Audio(shape=(None,), dtype=tf.int64),
    'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=7),
    'speaker_id': tf.string,
author = {Vlasenko, Bogdan and Schuller, Bjorn and Wendemuth, Andreas and Rigoll, Gerhard},
year = {2007},
month = {01},
pages = {2249-2252},
title = {Combining frame and turn-level information for robust recognition of emotions within speech},
journal = {Proceedings of Interspeech}