X-vector anonymization using autoencoders and adversarial training for preserving speech privacy | Publicación