As a guest user you are not logged in or recognized by your IP address. You have
access to the Front Matter, Abstracts, Author Index, Subject Index and the full
text of Open Access publications.
In this work, the authors propose a method for re-extracting deep learning features. By the approach of “Separation Statistics”, one can obtain a criterion of the principle of “Intraclass Aggregation and Interclass Dispersion”, so as to realize the re-extraction of the “deep learning” features. The re-extraction algorithm is applied to speaker recognition, and the experimental results indicate that the re-extracted deep learning features can effectively improve the accuracy of the speaker recognition. The findings reveal a fact that, although one can obtain more information depend on the deep learning, not all of the “more abstract” or “more deeper” features have positive effect on the task of recognition and classification, one should focus on the most important features rather than more features.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.