Speech signal recovery using an optimized multi-channel adaptive noise canceller based on bone- and air-conducted measurements

2021 Fiscal Year Final Research Report

• Project/Area Number: 18K04175
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 21030:Measurement engineering-related
• Research Institution: Prefectural University of Hiroshima
• Principal Investigator: XIAO YEGUI 県立広島大学, 地域創生学部, 教授 (50252325)
• Project Period (FY): 2018-04-01 – 2022-03-31
• Keywords: 音声復元 / 適応ノイズキャンセラー / 骨導音 / 気導音 / 線形結合器 / 非線形フィルタ / 最適化

Outline of Final Research Achievements

In this project, speech denoising is attempted by using of active noise canceller (ANC) that takes simultaneously the bone- and air-conducted (BC, AC) speech measurement as its reference and primary signal, respectively. Focus is put onto the use of multiple reference and/or primary channels. The main outcome is as follows: 1) an ANC with multiple BC signals is proposed, that improves the recovery quality as the number of reference BC is increased, 2) an ANC with multiple optimized reference channels is proposed, that the reference channels are connected by a constrained linear combiner (LC) which is optimized based on the minimization of the ANC error, and the recovery accuracy is further improved, and 3) we also tried an ANC that has multiple optimized primary (AC) channels and found that the recovery quality may also be improved. All proposed systems have been applied to real measurements to confirm their performance and capabilities.

Free Research Field

知能情報システム

Academic Significance and Societal Importance of the Research Achievements

本研究は以下の学術的意義や社会的意義があると考える
①　マルチチャンネルANCの最適化による音声復元は、初の試みとして、理論的貢献のみならず、音声復元の精度向上を実現し、その実用化へ大きく前進した。
②　骨導音・気導音を活用した製品化（システムを含む）が進む中、本研究は関連産業に刺激を与え、技術向上へ貢献できると考える。