Estimation of posture and joint angle of human body using foot pressure distribution: Morphological computation with human foot

TL;DR

This work addresses estimating upper-body posture and multi-joint angles from a single plantar sensor by measuring foot pressure distribution. It uses a simple, interpretable approach: Ridge regression with feature selection (correlation threshold $>0.15$) and a fixed hyperparameter $\\lambda = 10$, trained on squat data from seven adult males, with training/validation split. Results show $R^2$ around $0.9$ and RMSEs in the reported ranges, and perturbing plantar morphology (rubber or plastic interposers) significantly reduces accuracy, highlighting the plantar morphology's role and supporting the framing of the foot as a computational resource in morphological computation and physical reservoir computing. The study suggests practical plantar-only sensing for gait/posture monitoring and contributes to the conceptual bridge between biomechanics and unconventional computation, offering a low-cost, surface-level sensing paradigm with broad potential applications.

Abstract

This paper proposes a novel contact and wearable sensing system for estimating the upper body posture and joint angles (ankle, knee, and hip) of the human body using foot pressure distribution information obtained from a sensor attached to the plantar region. In the proposed estimation method, sensors are installed only on the plantar region, which is the end of the human body and the point of contact with the environment. The posture and joint angles of other parts of the body are estimated using only this information. As a contact and wearable sensor, the proposed system differs from previous measurement systems in the sense that the sensor does not need to be placed near the target joint or body. The estimation was carried out using a multivariate linear regression model with the foot pressure distribution as the input and the joint angle or posture as the output. The results reveal that it is possible to estimate the posture and joint angles of the human body from foot pressure distribution information (R2$\fallingdotseq$0.9). The proposed estimation method was validated by morphological computation to confirm that it is enabled by foot morphology. The validation approach compared the estimation accuracy achieved when an object was interposed between the foot pressure distribution sensor and the plantar region and the morphological relationship of the plantar region to the environment varied. The results reveal that there is a significant difference in the estimation accuracy between cases with and without an intervening object, suggesting that the morphology of the plantar region contributes to the estimation. Furthermore, the proposed estimation method is considered as physical reservoir computing, wherein the human foot is used as a computational resource.

Figures (4)

• Figure 1: Experimental setup: (a) (a-1) experimental setup in this study; (a-2) motion capture was used to obtain the posture and joint angles of the human body; (a-3) a sheet-shaped distributed pressure sensor was used to measure the pressure distribution on the plantar surface of the foot. (b) The experiment was conducted under three conditions: (b-1) condition wherein nothing was placed between the sole and the pressure distribution sensor (Condition (A): nothing); (b-2) condition wherein a rubber plate was placed between the sole and the pressure distribution sensor (Condition (B): rubber); (b-3) condition wherein a plastic plate was placed between the sole and the pressure distribution sensor (Condition (C): plastic). The experimental participants were equipped with motion-capture markers and performed natural squat exercises on a pressure distribution sensor. Foot pressure distribution data and motion capture data were collected during the exercise.
• Figure 2: Representative example of estimation results for: (a) ankle joint angle, (b) knee joint angle, (c) hip joint angle, and (d) upper body posture under experimental condition (A). This figure shows only the range of validation data. For comparison, the plots for each output (a)–(d) are overlaid with the time series of the measurement data $\theta$ (red line) and estimated data $\hat{\theta}$ (blue line). These results reveal the high estimation accuracy for the joint angles and posture using the foot pressure distribution.
• Figure 3: Results for error (RMSE and R2) in estimation of (a) ankle joint angle, (b) knee joint angle, (c) hip joint angle, and (d) upper body posture under each experimental condition. The figure also shows the statistical analyses performed on the RSME and R2 results for the data obtained under experimental Condition (A) and (B), and the data obtained under experimental Conditions (A) and (C). The statistical significance was set to p$<$0.05. The difference in the measurement accuracy between cases with and without an intervening object is significant, with lower estimation accuracy for the rubber plate (Condition (B)) and plastic plate (Condition (C)) compared with the case without an intervening object (Condition (A)), which is true for all results except the RSME of body posture estimation.
• Figure 4: (a) Locations where mechanoreceptors are considered to be abundant in the plantar region and (b) representative example of map of absolute value of weight $\omega$ in multivariate linear regression model for each pressure sensor location. By comparing (a) and (b), the locations where the information weights in the estimation method are large, and the locations where there are many mechanoreceptors in the plantar region, are qualitatively consistent to some extent.