The influence of measurements and feature types in automatic micro-behavior recognition in meal preparation
Alberto Fornaser, Teruhiro Mizumoto, Hiroiko Suwa, Keiichi Yasumoto, M. De Cecco · IEEE Instrumentation & Measurement Magazine · 2018
What distinguishes modern technology from solutions of the past is the capability of automatic systems to interact with humans. However, the creation and the organization of such interaction are not simple since they involve different beings who understand each other in order to instantiate a collaborative environment. For this reason, automatic action recognition has been a widely analyzed topic since the last decade and will be one of the most challenging research areas for the technology of the future. In this work, the focus is on the recognition of atomic actions or, as they are known, micro-behaviors. These represent the sequence of short operations, usually characterized by durations of less than 1 s, that a human subject performs to achieve a complex activity, a macro-behavior. Here, the cooking macro-behavior and related micro-behaviors were considered.