Design of channel coding model for recognition of infrequent events in time series

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500215
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perception information processing/Intelligent robotics
• Research Institution: Hosei University
• Principal Investigator: ITOU Katunobu 法政大学, 情報科学部, 教授 (30356472)
• Co-Investigator(Kenkyū-buntansha): NISHIJIMA Toshihisa 法政大学, 情報科学部, 教授 (70211456) ; HIROTSU Toshio 法政大学, 情報科学部, 教授 (10378268)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 表現 / 逸脱 / 階層モデル / 音声 / 音楽 / 演技音声

Outline of Final Research Achievements

Recognition performance of statistic models fall for infrequent events. In this study, as significant infrequent events, expressions in acoustic media were focused on. Expressions are assumed as deviation from expected features by listeners in their situations or contexts. Based on this assumption, various expression corpora, such as violin performance, voice acting, and emphasis for presentation speech, were constructed. In conclusion, some class of expression was assumed to be purposely generated by high-skilled performers. Expressive deviation was infrequent, however repeatedly generated locally, such as in a single song or a single lecture. Such expression was well-described by hierarchical probabilistic models such as HDP-HMM.

Free Research Field

音声・音響・音楽情報処理