Object Recognition Using Flexible Tactile Sensor

• Item type: Symposium(1)
• 公開日: 2021-03-15
• タイトル: Object Recognition Using Flexible Tactile Sensor
• タイトル / 言語: en
• タイトル: Object Recognition Using Flexible Tactile Sensor
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: Kyushu Institute of Technology
• 著者所属: Kyushu Institute of Technology／Japan Society for the Promotion of Science
• 著者所属: Osaka University
• 著者所属: Osaka University
• 著者所属: Kyushu Institute of Technology
• 著者所属(英): en
• 著者所属(英): Kyushu Institute of Technology
• 著者所属(英): en
• 著者所属(英): Kyushu Institute of Technology / Japan Society for the Promotion of Science
• 著者所属(英): en
• 著者所属(英): Osaka University
• 著者所属(英): en
• 著者所属(英): Osaka University
• 著者所属(英): en
• 著者所属(英): Kyushu Institute of Technology
• 著者名: Shoshi, Tokuno; Yuichiro, Tanaka; Takumi, Kawasetsu; Koh, Hosoda; Hakaru, Tamukoh
• 著者名(英): Shoshi, Tokuno; Yuichiro, Tanaka; Takumi, Kawasetsu; Koh, Hosoda; Hakaru, Tamukoh
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: We propose an object recognition system based on tactile information obtained from a tactile sensor. Our tactile sensor is made of flexible materials and composed of three parts: silicon rubber, liquid metal, and a coil printed on a circuit board. The sensor is mounted on a robot hand to acquire the tactile information of grasped objects. The tactile information for object classification is learned by an echo state network (ESN). The tactile time series data acquired by the tactile sensor are fed into the ESN for training and classification. To determine whether the system can classify two objects with different hardness levels and two objects with similar colors which cannot be classified by an image recognition system, we conducted two experiments. The classification accuracy of two objects with different hardness levels was 90%, and that of two objects with similar colors was 100%.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: We propose an object recognition system based on tactile information obtained from a tactile sensor. Our tactile sensor is made of flexible materials and composed of three parts: silicon rubber, liquid metal, and a coil printed on a circuit board. The sensor is mounted on a robot hand to acquire the tactile information of grasped objects. The tactile information for object classification is learned by an echo state network (ESN). The tactile time series data acquired by the tactile sensor are fed into the ESN for training and classification. To determine whether the system can classify two objects with different hardness levels and two objects with similar colors which cannot be classified by an image recognition system, we conducted two experiments. The classification accuracy of two objects with different hardness levels was 90%, and that of two objects with similar colors was 100%.
• 書誌情報: Proceedings of Asia Pacific Conference on Robot IoT System Development and Platform; 巻 2020, p. 81-82, 発行日 2021-03-15
• 出版者 / 言語: ja
• 出版者: 情報処理学会