Abstract: Existing deep embedding learning based speaker verification (SV) methods suffer from performance degradation under domain shift conditions. This can be alleviated through unsupervised domain ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results