An Attribute-Aligned Strategy for Learning Speech Representation
This addresses privacy and bias issues in speech technology for applications like emotion recognition and speaker verification, but it is incremental as it builds on existing representation learning methods.
The paper tackles the problem of speech signals containing multiple personal attributes that can lead to privacy leaks or bias, by proposing an attribute-aligned learning strategy using a layered-representation variational autoencoder to derive identity-free representations for speech emotion recognition and emotionless representations for speaker verification. It achieves competitive performance on identity-free SER and better performance on emotionless SV compared to state-of-the-art adversarial methods on the MSP-Podcast corpus, while reducing the model and training needed for multiple privacy-preserving tasks.
Advancement in speech technology has brought convenience to our life. However, the concern is on the rise as speech signal contains multiple personal attributes, which would lead to either sensitive information leakage or bias toward decision. In this work, we propose an attribute-aligned learning strategy to derive speech representation that can flexibly address these issues by attribute-selection mechanism. Specifically, we propose a layered-representation variational autoencoder (LR-VAE), which factorizes speech representation into attribute-sensitive nodes, to derive an identity-free representation for speech emotion recognition (SER), and an emotionless representation for speaker verification (SV). Our proposed method achieves competitive performances on identity-free SER and a better performance on emotionless SV, comparing to the current state-of-the-art method of using adversarial learning applied on a large emotion corpora, the MSP-Podcast. Also, our proposed learning strategy reduces the model and training process needed to achieve multiple privacy-preserving tasks.