Naslov
Convolutional Neural Network Based Emotion Classification Using Speech Data

Autori
Vrebčević, Nikola ; Mijić, Igor ; Petrinović, Davor

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, prošireni sažetak, ostalo

Izvornik
Abstract Book of Third International Workshop on Data Science (IWDS 2018) / Lončarić, Sven ; Šmuc, Tomislav - Zagreb : Znanstveni centar izvrsnosti za znanost o podatcima i kooperativne sustave, 2018, 21-23

Skup
3rd International Workshop on Data Science (IWDS 2018)

Mjesto i datum
Zagreb, Hrvatska, 16.10.2018

Vrsta sudjelovanja
Poster

Vrsta recenzije
Domaća recenzija

Ključne riječi
speech ; emotion recognition ; emotion ; convolutional neural network ; recurrent neural network ; neural network

Sažetak
The human voice is a unique form of signal that is easily generated and understood. Its main purpose is a transformation of abstract thoughts into physical phenomena and vice versa. From the theory and everyday experience, it is known that verbal correspondence between people comprises information from both linguistic and paralinguistic dimensions. The emotional state of a person in many situations can be observed through paralinguistic dimension. The previous statement gives the motivation to explore in the direction of automatic methods for emotional states recognition. Modern techniques such as deep learning provide researchers with powerful tools which could be easily utilized in real-life applications. This research examines a model which comprises convolutional, recurrent and deep neural network as a single neural network which is used for emotion recognition from speech. The model is trained and evaluated on spectrograms that are calculated from utterances provided in RECOLA emotional database.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA