Towards forensic speaker identification in Spanish using triplet loss

Abstract

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 slices. Then we test how similar these vectors are for a given speaker and how dissimilar are compared with other speakers. Based on these metrics we propose the calculation of the Likelihood Radio which is a cornerstone for forensic identification.