Repository of the addiotanary material of the paper TinySV: Speaker Verification in TinyML with On-device Learning, Pavan, Mombelli, Sinacori, Roveri, 2023.
Details on the training of the keyword spotting module can be found in the folder kws.
Details on the training of the d-vector extractor module can be found in the folder d-vector-extractor, along with the experiment on the adaptive speaker verification module.
The publicly released dataset can be found in d-vector-extractor/dataset/output.
A video demo of the proposed application can be found here.