Joseph: phonetic-aware speaker embedding for far-field speaker verification

Performing speaker verification (SV) at a distance from the sound source is challenging because of the interference of noise and reverberation. In such a situation, incorporating phonetic information into speaker embeddings can help reduce the adverse effects of noise and reverberation. Inspired by this observation, we propose a Jointly optimized speaker-embedding and phonetic-matching (Joseph) framework to exploit phonetic content for far-field SV. The framework encourages the speaker embeddings to preserve phonetic information by matching the frame-based feature maps of a speaker embedding network with wav2vec's vectors. The intuition is that phonetic information can preserve low-level acoustic dynamics with speaker information and thus partly compensate for the degradation due to noise and reverberation. Results show that the proposed framework outperforms the standard speaker embedding on the VOiCES Challenge 2019 evaluation set and the VoxCeleb1 test set. This indicates that leveraging phonetic information under far-field conditions is effective for learning robust speaker representations.