Corpus development considering pop-noise balance for robust speaker verification systems

2018 Fiscal Year Final Research Report

• Project/Area Number: 16K16088
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing
• Research Institution: Tokyo Metropolitan University
• Principal Investigator: Shiota Sayaka 首都大学東京, システムデザイン研究科, 助教 (90705039)
• Research Collaborator: KIYA Hitoshi ; ONO Nobutaka
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: 話者認識 / 話者照合 / 声の生体検知 / なりすまし検出 / ポップノイズ検出 / 音素リスト

Outline of Final Research Achievements

This research focuses on the corpus development considering pop-noise balance for speaker verification systems. Recently, spoofing against speaker verification systems became serious problems, and many research groups start to research anti-spoofing countermeasures. We focus on a voice liveness detection (VLD) approach which is regarded as one of the anti-spoofing countermeasure. The voice liveness detection approach is to detect the inputs came from a genuine human or a loudspeaker. To improve the robustness of the VLD, the pop-noise detection has been proposed. It reported that the accuracies depend on the sentences. Therefore, the pop-noise balanced sentences are designed, and the new database is recorded by using the designed sentences. From the experimental results, the designed sentence can improved the performance of the VLD systems. It means that the speaker verification systems can protect from the spoofing attacks by using the VLD systems.

Free Research Field

音声信号処理

Academic Significance and Societal Importance of the Research Achievements

近年、音声対話システムやスマートスピーカーの普及により音声を入出力インタフェースとして機械と会話や命令をする機会が増えてきている。特に電話の発信やスケジュールの確認など個人的な情報を発声する場合、これを第三者が盗聴してなりすまし音声に使って個人情報を抜き出したり誤った操作を行わせることができてしまうという問題がある。そのためなりすまし音声の検出は重要課題であり、近年話者照合という声を使った生体認証技術に関する分野においても国内外、多くの研究機関によって研究が行われている。なりすまし検出の精度が向上し普及することで機械との音声対話の安全性も大幅に向上することが期待できる。