A Novel Smart Chair System for Posture Classification and Invisible ECG Monitoring

Abstract

In recent years, employment in sedentary occupations has continuously risen. Office workers are more prone to prolonged static sitting, spending 65–80% of work hours sitting, increasing risks for multiple health problems, including cardiovascular diseases and musculoskeletal disorders. These adverse health effects lead to decreased productivity, increased absenteeism and health care costs. However, lack of regulation targeting these issues has oftentimes left them unattended. This article proposes a smart chair system, with posture and electrocardiography (ECG) monitoring modules, using an “invisible” sensing approach, to optimize working conditions, without hindering everyday tasks. For posture classification, machine learning models were trained and tested with datasets composed by center of mass coordinates in the seat plane, computed from the weight measured by load cells fixed under the seat. Models were trained and evaluated in the classification of five and seven sitting positions, achieving high accuracy results for all five-class models (>97.4%), and good results for some seven-class models, particularly the best performing k-NN model (87.5%). For ECG monitoring, signals were acquired at the armrests covered with conductive nappa, connected to a single-lead sensor. Following signal filtering and segmentation, several outlier detection methods were applied to remove extremely noisy segments with mislabeled R-peaks, but only DBSCAN showed satisfactory results for the ECG segmentation performance (88.21%) and accuracy (90.50%).

Intelligent systems for sitting posture monitoring and anomaly detection: an overview

“…The use of K-nearest neighbor (KNN)-based classification systems have also been proposed in the literature [ 15 – 19 ]. This classifier is based on the construction of a multidimensional feature space where it is assumed that those data belonging to the same class have similar characteristics, and are therefore grouped into nearby clusters.…”

“…These postural diagnosis systems can be divided into two stages (Fig. 1 ): A first stage in which a postural monitoring is carried out, capturing postural data of interest, and a second stage in which statistical [ 13 – 23 ] or intelligent techniques [ 5 , 15 , 17 , 19 , 24 – 32 ], among others, are used to identify the user’s postural state.

Fig.

…”

IoT System for Real-Time Posture Asymmetry Detection

“…More recent systems employ a reduced number of sensors and achieve a higher success rate in the identification of the sitting posture. For example, a system for the combined monitoring of the sitting posture and the electrocardiographic activity was proposed very recently by Pereira et al [ 36 ], to recognize eight sitting positions by means of three load cells and monitor the heartbeat by means of conductive nappa leather dry electrodes, and the system proposed by Roh et al in 2019 [ 37 ] identifies with excellent classification accuracy six different postures by using four low-cost load cells.…”

“…In order to keep the cost low without impairing the efficiency, we employed the minimum number of sensors necessary to detect the posture asymmetry in an effective way, i.e., four sensors (as in [ 37 , 43 ]). The adoption of only three sensors (as in [ 36 ], where only one sensor is located in the front part of the seat) is, in our opinion, not sufficient for our goal of detecting the asymmetry in the left-right direction of both the front and the back part of the seat. Compared to other systems developed with the goal of recognizing a fixed number of standardized postures based on several different classification algorithms (e.g., pattern recognition and machine learning techniques), the system we have developed is aimed at the detection of the asymmetry of the sitting position along the left-right direction.…”

Exploring the electrical robustness of conductive textile fasteners for wearable devices in different human motion conditions

“…These advantages include increased comfort and ergonomy, which are a result of higher flexibility, stretchability and conformability to the human body shapes 3 . Moreover, e-textile wearables tend to be more lightweight and unobtrusive, giving rise to the concept of ''invisible'' sensor integration, even outside clothing 4 , 5 . Especially designed to be worn on the body, e-textile wearables present an ideal electronic interface to gather and process human physiological data in real time as well as to improve bodily functions such as blood circulation and wound healing 6 – 8 .…”