cs.LGApr 9, 2026

Smartwatch-Based Sitting Time Estimation in Real-World Office Settings

Authors: Olivia Zhang, Zhilin Zhang

Organizations: The Hockaday School, Dallas TX 75229, USA · Lumos Alpha, Dallas TX 75248, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Sedentary behavior poses a major public health risk, being strongly linked to obesity, cardiovascular disease, and other chronic conditions. Accurately estimating sitting time is therefore critical for monitoring and improving individual health. This work addresses the problem in real-world office settings, where signals from the inertial measurement units (IMU) on a smartwatch were collected from office workers during their daily routines. We propose a method that estimates sitting time from the IMU signals by introducing the use of rotation vector sequences, derived from Euler angles, as a novel representation of movement dynamics. Experiments on a 34-hour dataset demonstrate that exploiting rotation vector sequences improves algorithm performance, highlighting their potential for robust sitting time estimation in natural environments.

Figures & tables

Figure 1: Comparison between the conventional smartwatch-IMU pipeline and the proposed pipeline. (a) Typical paradigm in prior work for sitting classification/sitting-time estimation, in which features are extracted from IMU signals (e.g., accelerometer and/or gyroscope streams) and fed into a classifier/regressor. (b) Proposed paradigm: in addition to the standard IMU streams, the rotation-vector sequences (indicated by red dashed lines) are treated as an additional signal channel for feature extraction. Features from both branches are fused and then used for classification or regression.
MethodRecallPrecisionF1-scoreAccuracy
Proposed0.9640.9780.9710.958
ST-DT0.9050.9470.9250.894
ST-DT Improved0.9280.9380.9330.904
CNN-LSTM0.8960.9670.9300.902
CNN-LSTM Improved0.9210.9810.9500.930
VAE0.9160.9740.9440.922
Table 1: Performance comparison of the proposed method with ST-DT, CNN-LSTM, VAE, MaskCAE, and Conformer. The variants adapted to utilize rotation vector sequences are denoted as ST-DT Improved, CNN-LSTM Improved, VAE Improved, MaskCAE Improved, and Conformer Improved. Recall, precision, and F1-score are reported with respect to the sitting class.

Explore similar work

CardsList

  1. ArmPoser: Real-Time, Calibration-Free Arm Pose Estimation from Smartwatch IMU

    Sep 8, 2026Bishnu Dev, Vasco Xu, Xi-Aan Loh +3Wearable SensingHuman Pose Estimation

    Mohamed bin Zayed University of Artificial Intelligence · University of Chicago · Northwestern University

  2. Sedentary Behavior Classification for Wearable Sensors with a CNN-BiLSTM Model

    Aug 3, 2026Yuliang Chen, Weiwei Shi, Jingjing Zou +10Human Activity RecognitionWearable Sensing

    Halıcıoğlu Data Science Institute, University of California, San Diego, La Jolla, CA 92093, USA · Herbert Wertheim School of Public Health and Human Longevity Science, University of California, San Diego, La Jolla, CA 92093, USA · Department of Computer Science and Engineering, University of California, San Diego, La Jolla, CA 92093, USA +4

  3. Deep Learning for Sleep Heart Rate Estimation from Accelerometers: Toward Population-Scale Cardiac Insight Without Optical Sensors

    Oct 5, 2026Tanbin Islam Rohan, Pranjol Sen Gupta, Tanusree Debi +1TransformerWearable Sensing

    Department of Information Technology Kennesaw State University Kennesaw, GA, USA · School of Computing University of Georgia Athens, GA, USA