Proceedings of ACSEL2021, The 8th Asian Conference on Safety and Education in Laboratory January 10-15, 2022
A study on detecting unusual state in laboratory by machine learning of sounds
2022 年 13 巻 2 号 p. 33-35
詳細
2022 年 13 巻 2 号 p. 33-35
Most accidents have preliminary signs, and detecting deviations from the usual state can be an effective methodology to prevent accidents from occurring. However, since the actual conditions in laboratories are so diverse, it is difficult to uniformly define "usual states". In this study, we focus on the sound in the laboratory, and propose a method to model the continuously monitored acoustic data of the laboratory by a Gaussian Mixture Model (GMM) to define the usual state of the laboratory. Using this model, the unusual state as deviations from the usual state in the laboratory can be detected by extracting sounds with low log-likelihood.
