Study of speech signal amplitude and phase reconstructions for understanding sound environment

2019 Fiscal Year Final Research Report

• Project/Area Number: 19K21546
• Project/Area Number (Other): 18H06482 (2018)
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund (2019); Single-year Grants (2018)
• Review Section: 1002:Human informatics, applied informatics and related fields
• Research Institution: Tokyo Metropolitan University
• Principal Investigator: Wakabayashi Yukoh 首都大学東京, システムデザイン研究科, 特任助教 (80826462)
• Project Period (FY): 2018-08-24 – 2020-03-31
• Keywords: 位相信号処理 / 音声強調 / 雑音抑圧 / 音声区間検出

Outline of Final Research Achievements

This project tackled the challenges of clean speech reconstruction from noisy observation by using relationships between speech amplitude and phase features.
In general, previous studies have separately constructed amplitude- and phase-based noise reduction algorithms. In contrast, the principal investigator proposed new algorithm that integrates the two features and confirmed its superiority over separately-handled method. In addition, he applied the relationship to another audio signal processing such as voice activity detection and showed that new integration method achieves higher performance than only amplitude-based method.

Free Research Field

音響信号処理

Academic Significance and Societal Importance of the Research Achievements

本研究において得られた，雑音抑圧と音声区間検出に対する結果が示すことは以下の通りである．学術的意義としては，これまで別々に研究されてきた振幅特徴と位相特徴の関連を考慮し，統合的に取り扱うことが信号処理の性能を改善する上で重要であり，より高性能な信号処理アルゴリズムの構築に繋がることが確認できたことである．社会的意義としては，雑音抑圧や音声区間検出の高性能化に伴い，今後の情報社会において必須となる音声認識性能の向上や遠隔会話システムにおける，より円滑な会話の実現が達成できることが挙げられる．