Skip to content

martinigoyanes/speech-emotion-recognition

Repository files navigation

Dimensional and Discrete Emotion Recognition using Multi-task Learning from Acoustic and Linguistic features extracted from Speech

Requirements

nltk==3.6.1 matplotlib==3.3.3 numpy==1.19.5 audiofile==0.4.1 contractions==0.0.49 opensmile==2.0.0 seaborn==0.11.1 pandas==1.1.4 SpeechRecognition==3.8.1 scipy==1.5.4 Keras==2.2.4 torch==1.5.1 imblearn==0.0 pyAudioAnalysis==0.3.7 scikit_learn==0.24.2 xgboost==1.4.2

Abstract

The majority of research in speech emotion recognition focuses on the classification of discrete emotions either from acoustic features or text features. This thesis demonstrates that the dimensional representation of emotions is also very valuable and it shows its advantages over categorical emotions. The thesis proposes two different systems which both use bimodal features (text and acoustics) in order to recognize discrete and dimensional emotions. A sequential system that first performs dimensional regression and then classification and a parallel system that performs classification and regression at the same time.

The thesis develops a multi-task regression model that serves as the core for both systems. Using the Concordance Correlation Coefficient (CCC) for evaluation it is discovered that the thesis developed architecture for dimensional regression outperforms across all dimensions (valence, arousal, dominance) the regression model introduced in previous research at the Cambridge institution. In addition, the thesis proves that the sequential system outperforms the parallel system in the recognition of both discrete (classification accuracy) and dimensional emotions (CCC). This finding verifies the validity of the theoretical model of psychology about dimensional emotions and its ability to represent a discrete emotion in a three dimensional space without losing any information. Furthermore, it is demonstrated how transfer learning can be used in this specific task to improve the results of the system.

Models

Regression Regression & Classification

Systems

Results

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published