Cooking Event Detection from Temporal Thermal Condition of Residential Home
Naima Khan, Nirmalya Roy · 2020
Contact-free activity detection is being used in several domains i.e., healthcare, cyber physical systems for its non-intrusive and flexible characteristics for end users. Thermal condition of residential homes are affected by both the outdoor weather conditions and the inside human activities. The activity of cooking affects the thermal comfort of residents inside the home and incurs a significant amount of electricity consumption in commercial kitchens. Though camera and body sensor based frameworks are proposed in the existing literature to detect cooking activities, contact free activity inference is necessary to non-intrusively assess the daily thermal comfort, monitor building envelope, electricity usage. In this work, we collected thermal condition of an apartment from surface temperature sensors and cooking activity of residents were recorded by visual observations. We avoided other activities i.e., opening doors or windows which can affect the thermal condition of residential home during cooking. We proposed a framework for cooking activity detection using recurrent neural network from surface temperature and humidity signals of residential homes. Our proposed algorithm achieves approximately 93% accuracy in cooking event detection from ambient building thermal condition.