We study the use of embedded and worn sensors to unobtrusively detect the activities of daily living (ADL). Our aim is to find the minimum set of sensors required to detect these basic tasks. In this exploratory work, we analyze the publicly available 'Intense Activity' dataset from the MIT PlaceLab project and study the classification of eating and meal preparation vs. other activities. We find that eating and meal preparation can be detected with an accuracy of 90% using less than 1/3 of the over 300 available sensors in the PlaceLab. If only 8 sensors are used, the accuracy is 82% which may be adequate for some applications.