Publications
JOURNAL (INTERNATIONAL) Grip Strength Estimation Using Input Data From a Commodity Smartphone: Model Development and Validation Study
Komei Tajima (Keio University), Kaori Ikematsu, Toshiya Isomoto, Kunihiro Kato (Tokyo University of Technology), Yuta Sugiura (Keio University)
JMIR Human Factors (JHF)
April 23, 2026
Background:
Grip strength is a crucial indicator of muscle deterioration, recovery, sarcopenia, and neurological disorders. However, conventional measurement requires a dedicated dynamometer, which limits accessibility and requires specific movements.
Objective:
This study aimed to propose and validate a method for estimating grip strength using standard smartphone operations, thereby eliminating the need for specialized equipment.
Methods:
Data were collected from 21 young adults in the main experiment, who performed standard smartphone tasks (tapping, flicking, and dragging) after measuring their grip strength with a dynamometer. A predictive regression model was developed using touch and inertial sensor data. The model was first evaluated using a random split of the entire dataset (random split evaluation). To further assess practical feasibility and generalizability, we conducted leave-one-user-out validation and a few-day calibration validation; the latter simulated a calibration scenario by incorporating 1 to 4 days of user-specific data into the training set.
Results:
The regression analysis using a random split of the dataset demonstrated high accuracy, with a mean absolute error of 2.62 (SD 0.18) kg, a mean absolute percentage error (MAPE) of 8.91% (SD 0.57%), and a coefficient of determination of 0.802 (SD 0.036). In the validation of practical scenarios, the leave-one-user-out validation resulted in a MAPE of 15.08% (SD 5.40%). However, the personalized few-day calibration model showed significant improvements as calibration days increased, with the MAPE decreasing to 13.96% (SD 5.57%) after 1 day and reaching 11.64% (SD 5.80%) after 4 days. Furthermore, the National Aeronautics and Space Administration Task Load Index assessment indicated a low overall subjective workload (mean 3.04, SD 2.23 on a scale of 10), confirming the method’s suitability for daily use without a significant burden on users.
Conclusions:
The proposed method demonstrates that smartphones can serve as a viable, pervasive tool for daily grip strength monitoring, offering a convenient alternative to traditional dynamometers.
Paper :
Grip Strength Estimation Using Input Data From a Commodity Smartphone: Model Development and Validation Study
(external link)