Speech Emotion Recognition with Dual-Sequence LSTM Architecture

[ICASSP 2020] This is an implementation of Speech Emotion Recognition with Dual-Sequence LSTM Architecture (DS-LSTM)

• Dual-Sequence LSTM (DS-LSTM)
• Dual-level model with DS-LSTM cell

Requirements

• Python 3
• PyTorch 1.0

Results

• IEMOCAP

	WA	UA
MDS-LSTM	69.4	69.5
MLSTM + MDS-LSTM	72.7	73.3

Acknowledgement

Jianyou Wang
Michael Xue
Ryan Culhane
Enmao Diao
Jie Ding
Vahid Tarokh