Fluid intake recognition using inertial sensors

Fluid Intake Monitoring Systems for the Elderly: A Review of the Literature

“…Several wearable techniques are proposed to detect food or liquid intake including inertial measurements of the wrist or body [ 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67 , 68 , 69 , 70 , 71 , 72 , 73 ], or textile-based measurements [ 74 , 75 , 76 , 77 , 78 , 79 , 80 , 81 , 82 , 83 , 84 , 85 , 86 ]. The majority of studies analyzed and classified food intake with less focus on liquid intake.…”

“…Several research papers have used Inertial Measurement Units (IMUs) on the wrist to measure liquid intake. Most previous work focused on intake recognition, but did not attempt to provide an intake volume estimate [ 45 , 46 , 47 , 48 , 49 ]. The wrist-mounted inertial sensors are often used in combination with machine learning (ML) algorithms to detect drinking events.…”

“…The majority of recent papers used Artificial Intelligence (AI) to detect drinking using wearable devices. Considering single wrist inertial devices, fluid intake classification was performed using Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) networks or LSTM alone [ 46 , 52 ], hierarchical temporal memory (HTM) [ 48 ], log regression with Artificial Neural Networks (ANN) [ 53 , 54 ], Random Forest (RF) [ 47 , 54 ], Hidden Markov Methods (HMMs) with k-nearest neighbors (kNN) [ 55 ] and HMMs with Gaussian Mixture Modelling (GMM) [ 56 ]. Weiss et al tested several ML algorithms on various Activities of Daily Living (ADL), including but not specific to drinking actions.…”

“…From the studies including a Null class, Wellnitz et al achieved an average accuracy of 95.7% for drinking detection using a single IMU on the wrist and 81.4% when placing the IMU on eye glasses [ 46 ]. Moreover, classical algorithms including Support Vector Machine (SVM), kNN and RF performed as well as the hybrid CNN-LSTM deep learning model [ 46 ].…”

“…From the studies including a Null class, Wellnitz et al achieved an average accuracy of 95.7% for drinking detection using a single IMU on the wrist and 81.4% when placing the IMU on eye glasses [ 46 ]. Moreover, classical algorithms including Support Vector Machine (SVM), kNN and RF performed as well as the hybrid CNN-LSTM deep learning model [ 46 ]. This study also found that there was a low correlation (0.49) between length of drink and volume consumed across the 41 participants, showing that this is not a reliable way to estimate volume intake [ 46 ].…”