楽音と音声の双方に適用できる客観品質推定尺度の検討

摘要

全IP（Internet Protocol）技術を特徴の一つとする次世代ネットワーク（NGN：Next Generation Network）における音サービスの主たる品質要因は，時間連続的な符号化ひずみ，時間離散的なパケット損ひずみ，及び信号の周波数帯域制限ひずみである．筆者らは，これまで楽音信号に対して，連続ひずみ，離散ひずみ，帯域制限ひずみを考慮した客観品質評価法を検討してきた[1]．本稿では，同じ構造の客観品質推定尺度を用いて，音声信号の客観品質推定について検討し，楽音と音声の双方に適用できる客観品質評価尺度の開発を目指す．%The main quality factors of sound services for Next Generation Network (NGN), one characteristic of which is all IP technologies, are continuous time codec distortion, discrete time packet loss distortion, and restricted frequency band distortion. Previously, authors have proposed objective quality assessment methods for music signals taking into consideration continuous time distortion, discrete time distortion, and restricted frequency band distortion for music signals. This paper examines the objective quality estimation for speech signals through the use of objective quality measures of the same structure and aims toward the development of objective quality measures applicable for both music and speech audio.