On the use of Self-supervised Pre-trained Acoustic and Linguistic Features for Continuous Speech Emotion Recognition
This work provides a strong performance improvement for continuous speech emotion recognition, a task often limited by small amounts of labeled data, benefiting researchers and developers in speech processing.
This paper explores the use of self-supervised pre-trained acoustic (wav2vec) and linguistic (camemBERT) features for continuous speech emotion recognition. The authors achieved a concordance correlation coefficient (CCC) of 0.825 on the AlloSat dataset, significantly outperforming the 0.592 CCC obtained with MFCC and word2vec.
Pre-training for feature extraction is an increasingly studied approach to get better continuous representations of audio and text content. In the present work, we use wav2vec and camemBERT as self-supervised learned models to represent our data in order to perform continuous emotion recognition from speech (SER) on AlloSat, a large French emotional database describing the satisfaction dimension, and on the state of the art corpus SEWA focusing on valence, arousal and liking dimensions. To the authors' knowledge, this paper presents the first study showing that the joint use of wav2vec and BERT-like pre-trained features is very relevant to deal with continuous SER task, usually characterized by a small amount of labeled training data. Evaluated by the well-known concordance correlation coefficient (CCC), our experiments show that we can reach a CCC value of 0.825 instead of 0.592 when using MFCC in conjunction with word2vec word embedding on the AlloSat dataset.