Deep Discriminative Feature Learning for Accent Recognition
This work addresses the problem of accurate accent recognition for speech processing systems, which could benefit applications like speech assistants and transcription services by improving their performance on accented speech.
This paper tackles accent recognition, which is a more challenging problem than speaker identification due to the need for group-level accent features. The authors improved a deep speaker identification framework by using a Convolutional Recurrent Neural Network as a front-end encoder and integrating local features with a Recurrent Neural Network to create utterance-level accent representations. Their proposed network, without data augmentation, significantly outperformed the baseline system in the Accented English Speech Recognition Challenge 2020, with Circle-Loss achieving the best discriminative optimization.
Accent recognition with deep learning framework is a similar work to deep speaker identification, they're both expected to give the input speech an identifiable representation. Compared with the individual-level features learned by speaker identification network, the deep accent recognition work throws a more challenging point that forging group-level accent features for speakers. In this paper, we borrow and improve the deep speaker identification framework to recognize accents, in detail, we adopt Convolutional Recurrent Neural Network as front-end encoder and integrate local features using Recurrent Neural Network to make an utterance-level accent representation. Novelly, to address overfitting, we simply add Connectionist Temporal Classification based speech recognition auxiliary task during training, and for ambiguous accent discrimination, we introduce some powerful discriminative loss functions in face recognition works to enhance the discriminative power of accent features. We show that our proposed network with discriminative training method (without data-augment) is significantly ahead of the baseline system on the accent classification track in the Accented English Speech Recognition Challenge 2020, where the loss function Circle-Loss has achieved the best discriminative optimization for accent representation.