Unsupervised Representation Learning with Future Observation Prediction for Speech Emotion Recognition

Abstract

Prior works on speech emotion recognition utilize various unsupervised learning approaches to deal with low-resource samples. However, these methods pay less attention to modeling the long-term dynamic dependency, which is important for speech emotion recognition. To deal with this problem, this paper combines the unsupervised representation learning strategy -- Future Observation Prediction (FOP), with transfer learning approaches (such as Fine-tuning and Hypercolumns). To verify the effectiveness of the proposed method, we conduct experiments on the IEMOCAP database. Experimental results demonstrate that our method is superior to currently advanced unsupervised learning strategies.

References in corpus (7)

• Adam: A Method for Stochastic Optimization
• Efficient Estimation of Word Representations in Vector Space
• Auto-Encoding Variational Bayes
• Representation Learning with Contrastive Predictive Coding
• Theoretical Models of Learning to Learn
• Semi-Supervised Learning with Ladder Networks
• Towards Speech Emotion Recognition "in the wild" using Aggregated Corpora and Deep Multi-Task Learning

Cited by in corpus (5)

• Learning Speech Emotion Representations in the Quaternion Domain
• Analysis of Predictive Coding Models for Phonemic Representation Learning in Small Datasets
• Speech-XLNet: Unsupervised Acoustic Model Pretraining For Self-Attention Networks
• Contrastive Unsupervised Learning for Speech Emotion Recognition
• From Physics to Representation: Audio Learning with Synthetic Pre-training via Procedural Generation