Parameter Efficient Finetuning for Speech Emotion Recognition and Domain Adaptation

arXiv:2402.11747v1 Announce Type: cross
Abstract: Foundation models have shown superior performance for speech emotion recognition (SER). However, given the limited data in emotion corpora, finetuning all parameters of large pre-trained models for SER can be both resource-intensive and susceptible to overfitting. This paper investigates parameter-efficient finetuning (PEFT) for SER. Various PEFT adaptors are systematically studied for both classification of discrete emotion categories and prediction of dimensional emotional attributes. The results demonstrate that the combination of PEFT methods surpasses full finetuning …

abstract arxiv cs.lg cs.sd data domain domain adaptation eess.as emotion finetuning foundation overfitting paper parameters peft performance pre-trained models recognition speech speech emotion type