Multimodal speech emotion recognition Using Audio and Text

Project for HSE Deep Learning course: emotions recognition from audio and text.

Dataset

IEMOCAMP dataset with 4 largest emotion classes and united class "happy" and "excited". Preprocessed and splitted into train, test, validation as 8:1:1.

Models architectures

---

• ARE model: CNN

baseline/baseline_audio.py

• TRE model: GRU

baseline/baseline_text.py