@inproceedings{HiyamaSehrYoshidaetal., author = {Hiyama, Toru and Sehr, Armin and Yoshida, Takahiro and Hangai, Seiichiro}, title = {Text/Language-Independent Unknown Speaker Rejection Method Using LSP Codes}, series = {IEEE 9th Global Conference on Consumer Electronics (GCCE 2020): 13-16 Oct. 2020, Kobe, Japan}, booktitle = {IEEE 9th Global Conference on Consumer Electronics (GCCE 2020): 13-16 Oct. 2020, Kobe, Japan}, doi = {10.1109/GCCE50665.2020.9291835}, abstract = {Scam phone calls pretending relatives have become a major social problem. However, current countermeasure against such scam phone calls stays only enlightenment activities although there is a limited number of technical measures in telephones. In this study, we have proposed a text-independent unknown speaker rejection method which recognizes whether a known speaker or an unknown speaker on receiving a call. In this method, we used a line spectral pair (LSP) spectrogram as the feature and a convolutional neural network (CNN) as the identification engine. From the evaluated results, the proposed method recognized known/unknown speakers with the false acceptance rate (FAR) of 13.1\% in Japanese speech and 25.2\% in German speech, respectively.}, language = {en} }