|
Abstract
|
|
|---|---|
| Detecting and identifying emotions expressed in speech signals is a very complex task that generally requires processing a large sample siz·e to extract intricate details and match the diversity of human expression in speech. There is not an emotional dataset commonly accepted as a standard test bench to evaluate the performance of the supervised machine learning algorithms when presented with extracted speech characteristics. This work proposes a generic platform to capture and validate emotional speech. The aim of the platform is collaborativecrowdsourcing and it can be used for any language (currently, it is available in four languages such as Spanish, English, German and French). As an example, a module for elicitation of stress in speech through a set of online interviews and other module for labeling recorded speech have been developed. This study is envisaged as the beginning of an effort to establish a large, cost-free standard speech corpus to assess emotions across multiple languages. | |
|
International
|
Si |
|
Congress
|
IWINAC 2019 |
|
|
960 |
|
Place
|
Almería, España |
|
Reviewers
|
Si |
|
ISBN/ISSN
|
0302-9743 |
|
|
10.1007 /978-3-030-19591-5_16 |
|
Start Date
|
03/06/2019 |
|
End Date
|
07/06/2019 |
|
From page
|
149 |
|
To page
|
160 |
|
|
Proceedings of IWINAC 2019 (LNCS 11486) |