Online incremental classification resonance networks for human-robot interaction

Abstract

In human-robot interaction, classification is one of the most important problems, and it is essential particularly when the robot recognizes the surroundings and chooses a reaction based on a certain situation. Each interaction is different since new people appear or the environment changes, and the robot should be able to adapt to different situations during a brief interaction. Thus, it is imperative that the classification is performed incrementally in real time. In this sense, an online incremental classification resonance network (OICRN) is proposed to enable incremental class learning in multi-class classification with high performance online. In OICRN, a scale-preserving projection process is introduced to use the raw input vectors online without a normalization process in advance. Objects can be described in a hierarchical semantics, and people also perceive them in this way. It leads to the need for hierarchical classification in machine learning. Thus, an online incremental hierarchical classification resonance network (OIHCRN) is proposed to enable online incremental class learning in hierarchical classification. By the proposed scale-preserving projection and prior label appending process, OIHCRN reflects the class dependency between class levels and simultaneously normalizes the input vector online. To demonstrate the effectiveness of the proposed networks, experiments are carried out using benchmark datasets. To demonstrate the applicability, OIHCRN is applied to a multimedia recommendation system for digital storytelling. When a digital companion communicates with a user, meaning is delivered effectively by providing appropriate multimedia based on the conversation and the user's context. CNN-OICRN, an integrated network of the Convolutional Neural Network (CNN) for feature extraction and the OICRN for classification, is proposed for model-based online face identification and applied to a robotic system that learns human identities through human-robot interactions. It is verified that the robot can learn the identity of a new user through human-robot interaction and the newly learned knowledge can be reflected in the future interaction.