NOSE

Abstract

How we cook and prepare our food has an enormous impact on our health and well-being. Specific cooking methods, like deep-frying, are linked to obesity and the degradation of food nutrients, which contribute to various diseases and health issues. We present NOSE, a Novel Odor Sensing Engine, that passively and continuously monitors gas emissions in the kitchen area using an array of six metal oxide semiconductor (MOS) gas sensors and detects the occurrence of deep-frying. To evaluate NOSE, we collected sensor data from five foods (chicken, fish, beef, potato, and onion) cooked with three methods (deep-frying, grilling, and boiling) and three common frying oils (canola, corn, and soybean) in three different kitchens in a controlled manner. We demonstrate that NOSE can classify cooking by deep-frying with an average F1-score of 0.89. Based on the in-laboratory findings, we deployed NOSE in two different real-world households throughout a three-week period and successfully detected the occurrence of frying cooking with an average F1-score of 0.72, which is a promising result considering the relatively small number of data samples collected. Our results show the potential of using NOSE as an assistive dietary monitoring tool that periodically reports to users about their cooking habits.