Prototyping of environment sound recognition assist system by developing a sound identification method which can output recognition results as early as possible every moment with streaming processing

2018 Fiscal Year Final Research Report

• Project/Area Number: 16K00352
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent robotics
• Research Institution: Niigata University
• Principal Investigator: Mamoru Iwaki 新潟大学, 自然科学系, 准教授 (20262595)
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: 音認識 / 漸次的な認識法 / ベイズ推定 / ベイズ更新 / 環境音データベース / 感情音声 / 行動支援 / 骨導ヘッドフォン

Outline of Final Research Achievements

(1) We conducted a one-week-long investigation of indoor environmental sounds for each research subject. Sounds which had an influence on the behavior after the sounds were selected and recorded to construct a sound database. The database was labelled by sound impression based on hearing experiment.
(2) We developed a sound recognition method which can output recognition results every moment and update degree of belief about the result as the following part of the sounds arrives, which was examined in both emotional speech recognition and environment sound recognition experiments. According to computer simulations, it was confirmed that the method had the intended property.
(3) We examined bone-conduction headphones as a presentation device for information of occurring sounds. Because perceived sounds direction through bone-conduction headphones had a tendency to slip off, we developed a correction method.

Free Research Field

音声聴覚情報処理

Academic Significance and Societal Importance of the Research Achievements

自然な音声インターフェースのにとって、非言語情報の自動認識は重要な課題である。この認識は音の聴取と同時進行しなければならない。既存の手法は全収録後に認識といったバッチ処理的であるという欠点がある。本研究の特色は、この要件を満たす漸次的な認識手法を開発することである。その原型は、時々刻々と得られる短時間フレーム毎の認識をベイズ更新によって取り込みそれまでの認識結果を逐次更新していくことである。聴覚障がい者の行動支援機器の開発や、会話の進行に応じて適切な相槌を打ったり、話し手の感情を察して次の発話に備えたりすることができる、人間にとって自然な音声インターフェースの開発などに応用できる。