Fluid Intake Recognition using Inertial Sensors

As one of many uses of body-worn inertial sensors, health monitoring applications can have a significant impact on the quality of life for a user even with inexpensive consumer electronics. In this paper, we address fluid intake monitoring as an activity recognition problem and conduct a user-study with 41 participants. We show that while an approach with a wrist-mounted sensor outperforms an approach with a head-mounted sensor, they can both be considered viable options for such a system. Furthermore we compare the classification performance of a hybrid CNN-LSTM artificial neural network with simpler baseline classifiers.