Exploring Symmetric and Asymmetric Bimanual Eating Detection with Inertial Sensors on the Wrist

Thomaz, Edison; Bedri, Abdelkareem; Prioleau, Temiloluwa; Essa, Irfan; Abowd, Gregory D.

doi:10.1145/3089341.3089345

Cited by 21 publications

(21 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The majority of the studies were conducted exclusively in controlled laboratory settings ( n = 50, 72.5%; e.g., [55,65]), followed by exclusively free-living settings ( n = 15, 21.7%; e.g., [5,51]), with fewer being conducted in both settings ( n = 3, 4.3%; [2,3,7]). One study (1.4%) did not report the environment setting.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Assessing Eating Behaviour Using Upper Limb Mounted Motion Sensors: A Systematic Review

et al. 2019

View full text Add to dashboard Cite

Wearable motion tracking sensors are now widely used to monitor physical activity, and have recently gained more attention in dietary monitoring research. The aim of this review is to synthesise research to date that utilises upper limb motion tracking sensors, either individually or in combination with other technologies (e.g., cameras, microphones), to objectively assess eating behaviour. Eleven electronic databases were searched in January 2019, and 653 distinct records were obtained. Including 10 studies found in backward and forward searches, a total of 69 studies met the inclusion criteria, with 28 published since 2017. Fifty studies were conducted exclusively in laboratory settings, 13 exclusively in free-living settings, and three in both settings. The most commonly used motion sensor was an accelerometer (64) worn on the wrist (60) or lower arm (5), while in most studies (45), accelerometers were used in combination with gyroscopes. Twenty-six studies used commercial-grade smartwatches or fitness bands, 11 used professional grade devices, and 32 used standalone sensor chipsets. The most used machine learning approaches were Support Vector Machine (SVM, n = 21), Random Forest (n = 19), Decision Tree (n = 16), Hidden Markov Model (HMM, n = 10) algorithms, and from 2017 Deep Learning (n = 5). While comparisons of the detection models are not valid due to the use of different datasets, the models that consider the sequential context of data across time, such as HMM and Deep Learning, show promising results for eating activity detection. We discuss opportunities for future research and emerging applications in the context of dietary assessment and monitoring.

show abstract

Section: Resultsmentioning

confidence: 99%

“…The laboratory environment may affect the participant’s natural behaviour in the progress of experiment. Therefore, 10 studies (14.5%) conducted semi-controlled experiments in laboratory environments (i.e., [2,14,65]) or in a cafeteria, restaurant, or dining hall (i.e., [1,12,39,52,60,61,77]).…”

Section: Resultsmentioning

confidence: 99%

Assessing Eating Behaviour Using Upper Limb Mounted Motion Sensors: A Systematic Review

et al. 2019

View full text Add to dashboard Cite

show abstract

“…The authors report an average F1 score of 0.757 in their Leave-One-Subject-Out (LOSO) experiments using the most descriptive feature subset coupled with the AdaBoost classifier. Similarly, the recent work of Thomaz et al [32] also models the process of eating as a bimanual (i.e. two hand) task.…”

Section: Related Workmentioning

confidence: 99%

Modeling Wrist Micromovements to Measure In-Meal Eating Behavior From Inertial Sensor Data

Kyritsis

Diou

Delopoulos

2019

IEEE J. Biomed. Health Inform.

118

View full text Add to dashboard Cite

Overweight and obesity are both associated with in-meal eating parameters such as eating speed. Recently, the plethora of available wearable devices in the market ignited the interest of both the scientific community and the industry towards unobtrusive solutions for eating behavior monitoring. In this paper we present an algorithm for automatically detecting the in-meal food intake cycles using the inertial signals (acceleration and orientation velocity) from an off-the-shelf smartwatch. We use 5 specific wrist micromovements to model the series of actions leading to and following an intake event (i.e. bite). Food intake detection is performed in two steps. In the first step we process windows of raw sensor streams and estimate their micromovement probability distributions by means of a Convolutional Neural Network (CNN). In the second step we use a Long-Short Term Memory (LSTM) network to capture the temporal evolution and classify sequences of windows as food intake cycles. Evaluation is performed using a challenging dataset of 21 meals from 12 subjects. In our experiments we compare the performance of our algorithm against three state-of-the-art approaches, where our approach achieves the highest F1 detection score (0.913 in the Leave-One-Subject-Out experiment). The dataset used in the experiments is available at https://mug.ee.auth.gr/intake-cycle-detection/.

show abstract

“…Softmax layer function is used mainly in neural network in case of normalizing exponential when there are multiclass in order to describe and calculate to which category some input belongs according to the highest probability as (4). The purpose of the softmax classification layer is simply to transform all the DNN activations in the final output layer to a series of values that can be interpreted as probabilities.…”

Section: Deep Neural Network (Dnn)mentioning

confidence: 99%

“…The purpose of the softmax classification layer is simply to transform all the DNN activations in the final output layer to a series of values that can be interpreted as probabilities. So, the softmax function is applied onto the DNN outputs without an activation function or bias [35]: (4) where are the conditional probabilities in terms of the likelihood of event x occurring given that and C j ,respectively. , are the probabilities of observing respectively independently of each other.…”

Section: Deep Neural Network (Dnn)mentioning

confidence: 99%

Deep Neural Network for Human Activity Recognition

Khairy¹

2018

IJCA

View full text Add to dashboard Cite

Human activity recognition (HAR) has given a lot of attention in the recent years due to the need of high level context about the human activities in several applications. Many domains have attempted to overcome the lack of performance techniques used to collect raw data such as cameras to record or capture activities and inertial sensor units to record correct readings. As a result, few studies have regarded to acquire raw data and extract features instead of understanding, recognizing, inferring, and predicting human activities in future to obtain recommendations or detecting healthcare, daily, and educational positions to humans. This paper aims to analyze the performance of Neural Network (NN) and Deep Neural Network(DNN) for HAR. To achieve this aim, we select Daily and Sports Activities data set (DSA) to match paper's needs. This paper depends on NN and DNN based on softmax function. We form three sets of DSA dataset: small, medium, and large. The results showed that DNN based on softmax function reduce the computational cost than NN, increase the performance of network, and achieved high overall successful differentiation rate in testing on large dataset (97.74%) than on medium dataset (67.81%). or on small dataset (67.63%).

show abstract

Exploring Symmetric and Asymmetric Bimanual Eating Detection with Inertial Sensors on the Wrist

Cited by 21 publications

References 16 publications

Assessing Eating Behaviour Using Upper Limb Mounted Motion Sensors: A Systematic Review

Assessing Eating Behaviour Using Upper Limb Mounted Motion Sensors: A Systematic Review

Modeling Wrist Micromovements to Measure In-Meal Eating Behavior From Inertial Sensor Data

Deep Neural Network for Human Activity Recognition

Contact Info

Product

Resources

About