Leveraging Unlabeled Audio-Visual Data in Speech Emotion Recognition using Knowledge Distillation
This addresses the data cost issue for developers of human-computer interaction systems, but it is incremental as it applies an existing technique (knowledge distillation) to SER.
The paper tackles the problem of expensive labeled data collection for speech emotion recognition (SER) by proposing a knowledge distillation framework called LiSER that leverages unlabeled audio-visual data, reducing dependence on labeled datasets as demonstrated on RAVDESS and CREMA-D benchmarks.
Voice interfaces integral to the human-computer interaction systems can benefit from speech emotion recognition (SER) to customize responses based on user emotions. Since humans convey emotions through multi-modal audio-visual cues, developing SER systems using both the modalities is beneficial. However, collecting a vast amount of labeled data for their development is expensive. This paper proposes a knowledge distillation framework called LightweightSER (LiSER) that leverages unlabeled audio-visual data for SER, using large teacher models built on advanced speech and face representation models. LiSER transfers knowledge regarding speech emotions and facial expressions from the teacher models to lightweight student models. Experiments conducted on two benchmark datasets, RAVDESS and CREMA-D, demonstrate that LiSER can reduce the dependence on extensive labeled datasets for SER tasks.