Iván Meza

Towards forensic speaker identification in Spanish using triplet loss

2020

This work explores the use of a triplet loss deep network setting for the forensic identification of speakers in Spanish. Within the framework, we train a convolutional network to produce vector representations of speech spectrogram …