Improving the recognition accuracy and applicability of biometric authentication using Kernel Mutual Subspace Method

2014 Fiscal Year Final Research Report

• Project/Area Number: 25820155
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Communication/Network engineering
• Research Institution: The University of Electro-Communications
• Principal Investigator: ICHINO MASATSUGU 電気通信大学, 情報理工学(系)研究科, 助教 (80548892)
• Project Period (FY): 2013-04-01 – 2015-03-31
• Keywords: 生体認証 / 核非線形相互部分空間法 / 行動的特徴

Outline of Final Research Achievements

The biometric authentication based on behavioral characteristic is recently focused. Kernel mutual subspace method(KMS) is effective algorithm for biometric authentication based on behavioral characteristic because the distribution of behavioral characteristic is often nonlinear such as S-shaped curve. We studied three themes to improve the recognition performance of KMS. (1) We evaluated the effectiveness of KMS experimantally and discuss the reasons of effectiveness. (2) We evaluated the whether the characteristic that is effective for recognition is presenct in canonical angles. Based on the evaluation, we also study the fusion method of canonical angles for speaker recognition using lip movement. (3) We study the method to extract the common information to highlight the personal characteristics of each person. We studied the gender recognition method for extraction common information.

Free Research Field

セキュリティ