Bi-directional retrieval of speech and image by indexing both speech and image data.

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500124
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Media informatics/Database
• Research Institution: Iwate Prefectural University
• Principal Investigator: YOSHIAKI Ito 岩手県立大学, ソフトウェア情報学部, 教授 (90325928)
• Co-Investigator(Kenkyū-buntansha): OOKAWA Shigeki 千葉工業大学, 工学部, 教授 (40306395) ; TANAKA Kazuyo 筑波大学, 図書館情報メディア研究科, 教授 (70344207)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 画像、文章、音声等認識

Outline of Final Research Achievements

Subword recognition is performed for speech data streams, and we developed the new method to construct an index of syllable bigrams. When a user input query keywords, the method first selects a limited number of utterances including bigrams in query keywords, and enables to quickly identify the sections where query keywords are spoken in spoken documents
We also developed the method to search similar images quickly for a large amount of images and video data using Particle Swarm Optimization (PSO), and demonstrated the possibility of both directional search by the prototype system, "Searching/detecting an object using a speech"

Free Research Field

音声言語処理