Abstract: Adversarial attack approaches to speaker identification either need high computational cost or are not very effective, to our knowledge. To address this issue, in this letter, we propose a ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results