chandra380 / emospeech-dataset-v1

EMOSPEECH-DATASET version 0.1

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Data Description

folder name represents the location from where the audio is collected.

Format of this datset is described as below

keyword-environment-emotion-timestamp-hash.wav

example 1 :-

bacho-safe-angry-1562907046662-d14a028c2a3a2bc9476102bb288234c415a2b01f828ea62ac5b3e42f.wav

here, bacho is keyword, safe is environment, and angry as emotion

example 2 :-

background-safe-None-20190701193438-d14a028c2a3a2bc9476102bb288234c415a2b01f828ea62ac5b3e42f.wav

NOTE - Hash value d14a028c2a3a2bc9476102bb288234c415a2b01f828ea62ac5b3e42f represents Unknown user

possible values of keywords,environments,emotion is described below:

Keywords:

• help
• bacho
• stop
• no
• go
• yes
• unknown
• background

Emotions:

• Happy
• Angry
• Fearful
• Calm
• None

Environments:

• safe
• gunshots
• glass break

NOTE - background represents absense of keyword, it is not an actual keyword, and unknown represents keyword that are not present in list

Citation

Please consider citing.

@misc{1910.13801,
Author = {Subham Banga and Ujjwal Upadhyay and Piyush Agarwal and Aniket Sharma and Prerana Mukherjee},
Title = {Indian EmoSpeech Command Dataset: A dataset for emotion based speech recognition in the wild},
Year = {2019},
Eprint = {arXiv:1910.13801},
}