Fraunhofer Institute for Digital Media Technology IDMT
Cooking is part of everyday life – but risks lurk in the details: boiling over, grease fires, and unnecessary energy consumption. Previous solutions for monitoring cooking only record simple metrics such as water temperature and are therefore very limited in their perception and reaction capabilities. This leads to limited functionality and poor user-friendliness. This is where the EXIST project KitchenGuard, led by the interdisciplinary team of Jochen Heudorfer, Simon Heudorfer, and Jan-Philipp Bott, comes in: The goal is to develop an AI-supported cooking and frying sensor that analyzes cooking noises and precisely recognizes cooking conditions for each cooking zone, warns of critical situations at an early stage, and enables assistive functions for a comfortable cooking experience. This significantly increases safety and energy efficiency at the hob. The KitchenGuard team is now working with Fraunhofer IDMT to optimize the sensor technology.
Acoustic status monitoring of the cooking process
The basic idea behind the KitchenGuard sensor is based on the observation that people often judge the cooking process by the sounds they hear – when they cannot see the food directly. This human auditory perception is replicated using machine learning (ML) methods. This enables the KitchenGuard sensor to precisely detect the current cooking status based on the characteristic features of a recorded cooking sound.
Responsibilities of Fraunhofer IDMT
The involvement of Fraunhofer IDMT is intended to further improve cooking zone-specific sound detection, including for the simultaneous monitoring of multiple cooking pots. Fraunhofer IDMT is contributing its expertise in acoustic signal processing, sensor technologies, and machine learning methods to precisely detect the cooking status of each cooking zone. The main challenge is to separate the sound fields of the individual cooking zones from each other and from ambient noise as accurately as possible in order to optimally support downstream sound field analysis. To this end, Fraunhofer IDMT is developing a microphone concept for cooking zone-specific sound detection, testing integration options, and developing a microphone array that can be adapted to the installation situation as a populated printed circuit board with selected MEMS microphones. Digital signal preprocessing and a software tool are used to electronically adjust the recording properties (in particular the directivity) and tune them to the target cooking zones.