著者
佐々木 洋子 金吉 雅人 加賀美 聡 溝口 博 榎本 格士
出版者
一般社団法人日本機械学会
雑誌
ロボティクス・メカトロニクス講演会講演概要集
巻号頁・発行日
vol.2009, pp."1P1-C13(1)"-"1P1-C13(4)", 2009-05-25

This paper proposes a sound identification method for a mobile robot in home and office environment. We propose a simple sound database called Pitch-Cluster-Maps(PCMs) based on Vector Quantization approach. Binarized frequency spectrum is used for PCMs codebook generation. It can describe a variety of sound sources, not only voice, from short term sound input. The proposed PCMs sound identification requires several tens(msec) of sound input, and is suitable for a mobile robot application which condition is dynamically changing. We implemented the proposed method on our mobile robot audition system equipped with a 32ch microphone array. Robot noise reduction using proposed PCMs recognition is applied to each input signal of a microphone array. The performance of daily sound recognition for separated sound sources from robot in motion is evaluated.