Deep Multitask Learning for Pervasive BMI Estimation and Identity Recognition in Smart Beds

TL;DR

This work addresses the need for pervasive health monitoring by estimating BMI and recognizing user identity from pressure distributions captured by textile-based smart-bed sensors. It introduces a deep multitask neural network that learns a shared representation from 14 pressure-derived features to simultaneously perform BMI estimation and subject identification across multiple lying postures. Across two public datasets, the approach achieves state-of-the-art performance in both BMI prediction (high $R^2$ and low RMSE) and identity recognition, outperforming generative and discriminative baselines in 10-fold cross-validation. The results support deployment of smart-bed systems for automated health tracking and personalized ambient adjustments in smart homes and clinical settings.

Abstract

Smart devices in the Internet of Things (IoT) paradigm provide a variety of unobtrusive and pervasive means for continuous monitoring of bio-metrics and health information. Furthermore, automated personalization and authentication through such smart systems can enable better user experience and security. In this paper, simultaneous estimation and monitoring of body mass index (BMI) and user identity recognition through a unified machine learning framework using smart beds is explored. To this end, we utilize pressure data collected from textile-based sensor arrays integrated onto a mattress to estimate the BMI values of subjects and classify their identities in different positions by using a deep multitask neural network. First, we filter and extract 14 features from the data and subsequently employ deep neural networks for BMI estimation and subject identification on two different public datasets. Finally, we demonstrate that our proposed solution outperforms prior works and several machine learning benchmarks by a considerable margin, while also estimating users' BMI in a 10-fold cross-validation scheme.

Figures (10)

• Figure 1: An overview of the work presented in this paper is illustrated. Data was first collected using a smart bed goldberger2000physiobank, producing a set of pressure images. We then extracted features, which were fed into machine learning models to estimate BMI and identify subjects simultaneously.
• Figure 2: The $10$ common sleeping posture groups used for the study are presented. Participants assumed the postures while lying flat on a pressure sensing mattress. Posture images were reproduced from pouyan2017pressure.
• Figure 3: The effect of window size when applying the spatial filter is illustrated. The first image on the left presents the input unfiltered frame, while the next three present the impact of $2 \times 2$, $3 \times 3$, and $6 \times 6$ filter window sizes respectively.
• Figure 4: Examples of the calculated isolines of the pressure maps from PmatData is presented in this figure, where the definition of the lines provides two additional features.
• Figure 5: (a) The participants' BMI values for PmatData are illustrated, showing relatively unique measurements for each subject. (b) The t-SNE visualization of subjects over our feature space is illustrated, showing that each subject forms fairly separated clusters, indicating the potential for subject identification using the original feature space.