Study on probabilistic indicator for person-equivalence and anonymity

2015 Fiscal Year Final Research Report

• Project/Area Number: 26540089
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing
• Research Institution: The Institute of Statistical Mathematics
• Principal Investigator: Matsui Tomoko 統計数理研究所, モデリング研究系, 教授 (10370090)
• Co-Investigator(Kenkyū-buntansha): TAKEDA KAZUYA 名古屋大学, 情報科学研究科, 教授 (20273295) ; MINAMI KAZUHIRO 統計数理研究所, モデリング研究系, 准教授 (10579410)
• Project Period (FY): 2014-04-01 – 2016-03-31
• Keywords: プライバシー保護 / 個人認証 / 音声データ

Outline of Final Research Achievements

We studied on maximum mean discrepancy (MMD) which is an instance of an integral probability metric with kernel methods and found that it is difficult to use MMD as the probabilistic indicator for person-equivalence for non-i.i.d. data. To deal with the problem, we examined to use the wild bootstrap method and found that there were dynamic and strong correlation in speech data and the high performance of the wild bootstrap method was not expected. Morever, we experimentally found that a deep neural network (DNN) based method did not perform well when the available data amount was not sufficient. Future work includes to investigate a new research direction for the probabilistic indicator for person-equivalence.

Free Research Field

統計的機械学習、音情報処理