What the Sleeve Feels: Explainable Machine Learning for Textile Pressure-Based Postural Screening

Limon Bin Hossain; Md Sadib Rahman Ananta

cs.AIOct 7, 2026

What the Sleeve Feels: Explainable Machine Learning for Textile Pressure-Based Postural Screening

Authors: Limon Bin Hossain, Md Sadib Rahman Ananta

Organizations: Department of Industrial and Production Engineering, Bangladesh University of Engineering and Technology (BUET), Dhaka, 1000, Bangladesh

Abstract

Pressure-sensing smart textiles convert body-surface contact into a dense, image-like signal closely tied to posture and movement, making them a promising low-cost route to wearable posture screening. Realizing that promise, however, requires more than classification accuracy: a deployable system must generalize to wearers unseen during training, expose the physical evidence behind its decisions, and tolerate the small donning offsets that occur whenever a garment is removed and re-worn. This paper addresses these three requirements jointly using a knitted piezoresistive sleeve worn on the forearm as a testbed. We regroup fine-grained everyday activities into three coarser screening categories (neutral, potentially undesirable, and functional or transitional), engineer 29 interpretable pressure-distribution features spanning global intensity, spatial center of pressure, quadrant asymmetry, distribution complexity, and short-horizon temporal change, and evaluate under a strict subject-wise split. A tuned XGBoost classifier reaches 0.818 accuracy, 0.788 balanced accuracy, and 0.801 macro F1 on unseen test subjects, with tight frame-level bootstrap 95% intervals of about plus-minus 0.01 and a subject-to-subject standard deviation near 0.06 under leave-one-subject-out cross-validation. A simple 2D-CNN baseline trained on raw frames achieves broadly similar performance, showing that hand-engineered features are not left behind by a learned spatial representation on this task. SHAP-based explanation, a feature-group ablation, per-activity error analysis inside the pooled undesirable class, class-mapping sensitivity, and a simulated donning-rotation stress test together locate what the model relies on, where it degrades, and why, directly targeting the generalization, interpretability, and robustness gaps that determine whether such a system is deployable.

Figures & tables

Fig. 1: Representative photographs and corresponding pressure maps for the 18 everyday activities in the Smart-Sleeve dataset. Image reproduced from the Smart-Sleeve dataset repository [ 23 ] .

Fig. 2: Representative raw $20\times 10$ pressure frames for eight activities spanning the three screening categories.

Model	Acc.	Bal. Acc.	Macro F1	ROC AUC
Logistic Regression	0.730	0.668	0.681	0.873
Random Forest	0.793	0.737	0.762	0.933
XGBoost	0.818	0.788	0.801	0.942
TinyCNN (raw frames)	0.814	0.779	0.795	—

TABLE I: Test-set performance on unseen subjects (subjects 10, 12; 12,506 frames)

Fig. 3: Row-normalized confusion matrix for XGBoost on the held-out test subjects.

Subj.	Frames	Acc.	Bal. Acc.	Macro F1
10	6,752	0.832	0.822	0.828
12	5,754	0.802	0.744	0.763

TABLE II: Per-test-subject XGBoost metrics on the two held-out subjects. Subject 10’s numbers are close to the aggregate test figures, while subject 12’s balanced accuracy and macro F1 sit below them, foreshadowing the subject-level spread quantified by leave-one-subject-out cross-validation.

Activity	Frames	Recall	% N	% U	% F
Lean back at work	661	0.735	17.7	73.5	8.8
Sit with arms on the legs	844	0.829	0.0	82.9	17.1
Lean forward at work	710	0.856	14.4	85.6	0.0
Play mobile phone	749	0.944	0.5	94.4	5.1
Write with a hunchback	834	0.946	5.4	94.6	0.0
Sleep on the table	723	1.000	0.0	100.0	0.0

TABLE III: Per-activity behavior of XGBoost inside the pooled potentially undesirable class on the test subjects.

Fig. 4: Leave-one-subject-out macro F1 for XGBoost, one bar per held-out subject; red dashed line marks the mean.

Feature group	# feats	Acc.	Bal. Acc.	Macro F1
All	29	0.818	0.788	0.801
Spatial	6	0.688	0.639	0.643
Global intensity	9	0.629	0.569	0.574
Asymmetry	6	0.582	0.539	0.539
Complexity	3	0.505	0.511	0.495
Temporal	2	0.486	0.447	0.449

TABLE IV: Feature-group ablation on the held-out test subjects.

Mapping	Acc.	Bal. Acc.	Macro F1
Original	0.818	0.788	0.801
Alt. A (borderline swap)	0.841	0.790	0.810
Alt. B (stricter undesirable)	0.827	0.815	0.815

TABLE V: Test-set XGBoost performance under three three-class mappings evaluated on the same held-out subjects.

Fig. 5: SHAP beeswarm plot for the potentially undesirable class. Each point is one test-set frame; horizontal position is the feature’s SHAP value and color encodes the feature’s own numerical value from low (blue) to high (red).

Shift (taxels)	Acc.	Bal. Acc.	Macro F1
$-3$	0.533	0.489	0.471
$-2$	0.614	0.541	0.522
$-1$	0.761	0.706	0.721
$0$	0.818	0.788	0.801
$+1$	0.807	0.775	0.787
$+2$	0.749	0.719	0.728

TABLE VI: Robustness of the frozen XGBoost model to simulated circumferential donning rotation.

Explore similar work

Sep 16, 2026cs.LG

Smart Insole Human Activity Recognition for Continuous Monitoring in Elderly Care

Falls in older adults are often preceded by changes in mobility, balance, and postural transitions. This paper presents a wireless smart insole platform and machine-learning workflow for recognizing sitting, standing, walking, and unstable walking from plantar-pressure and inertial signals. Each insole integrates 16 active pressure-sensing locations and a six-dimensional IMU stream consisting of tri-axial acceleration and angular velocity. Data were collected from 15 healthy adults at 80~Hz and segmented into overlapping windows. Window length and candidate model families were first screened with stratified 10-fold cross-validation; the primary performance estimate was then obtained with participant-independent 5-fold Stratified Group cross-validation, ensuring that all windows from a participant remained in a single fold. Under this protocol, Histogram-Based Gradient Boosting (HGB) achieved macro-F1 scores of 0.954 and 0.959 for the left and right feet, respectively, and 0.980 with bilateral sensing. A compact 1D-CNN evaluated with the same participant-independent folds did not significantly outperform HGB (

p=0.0625

). The results show that low-profile footwear sensing can infer activity state from pressure and IMU measurements for participants unseen during training, establishing a basis for activity monitoring and fall prevention in elderly care.

Edwin Rios, Antony Garcia, Fengpei Yuan +1

Human Activity Recognition Inertial Measurement Units Garments Cross Validation

Ask PDF

Aug 4, 2026cs.LG

Tactus: Open-Vocabulary Object Recognition from Low-Cost Pressure Arrays

Resistive pressure arrays are the cheapest and most widely shipped tactile sensors, yet tactile representation learning has concentrated on optical sensors that image a deforming gel. We present Tactus, an open model that answers text queries from pressure data alone: on the STAG benchmark (27 objects, held-out recordings), it reaches 0.771 +/- 0.062 top-1 over four runs (top-3 0.935), matching, and at best exceeding, the dataset's supervised closed-set CNN at 0.76, with no trained classifier head. The recipe is small-data: 187 training recordings, masked-autoencoder pretraining on 144k unlabeled same-sensor frames, and the sensor's own calibration affine, which recovered more accuracy than every architecture change combined. The released model's errors concentrate in a few contact-ambiguous classes, are uncorrelated with text-target geometry (Spearman rho <= 0.05 over 702 class pairs), and survive paraphrased and even bare-name queries within one point; two diverse frames recover 89% of eight-frame accuracy. Failures are reported with equal precision: cross-sensor pretraining pooling gave no gain, vision co-training degraded touch, and a mis-normalized input pipeline silently discarded 97% of the sensor's dynamic range while producing plausible intermediate results. Weights, code, and the memory layer the model plugs into are released openly.

Abdul Basit Tonmoy

Eximius Labs Wabash College

Tactile Hand Gesture Recognition Open-Set Recognition Geometric Constraints

Ask PDF

Sep 15, 2026cs.HC

Beyond Gestures: Estimating Full Hand Pose and Contact Forces from Wrist-Worn Pressure Sensor Array

Capturing hand motion and interaction forces is critical for interactive computing, VR, and high-fidelity tactile demonstrations for robot learning. We introduce a wrist-worn pressure-sensing wristband that recovers continuous full-hand pose and distributed contact force on a single wearable. The system consists of flexible capacitive sensor arrays around the wrist, which require no electrical skin contact, and a recurrent network that maps the resulting pressure signal to hand state. Our key insight is that muscle contraction and tendon displacement produce pressure patterns, which correlate strongly with hand pose and interaction force. To validate this, we collect synchronized recordings of wrist pressure, optical motion-capture hand pose, and tactile-glove interaction force, covering isolated finger motion, fingertip-force stress tests, and natural hand-object manipulation. On isolated single-user motion the wristband attains

4.6^\circ

mean finger-joint MAE, and across four users manipulating everyday objects it estimates per-finger contact force at

R^2=0.57

, which an external pose signal brings up to

0.75

. We see the wristband as one node in a constellation of everyday wearables -- e.g. paired with an egocentric camera -- adding the contact force that vision cannot observe and taking over when the hand is occluded.

Svetoslav Kolev, Lingni Ma, Michael Goesele +3

Meta Reality Labs Research

Finger Ground Reaction Forces 3D Hand Hand Force

Ask PDF