ChairPose: Vision-Free Seated Pose Estimation
- ChairPose is a wearable-free, vision-free system that uses pressure sensing combined with explicit chair geometry to accurately reconstruct full-body 3D seated pose.
- It employs a two-stage generative model—MotionQuantizer and Pressure2Pose—that tokenizes motion and autoregressively decodes pressure data into 3D joint coordinates.
- The system enhances performance across diverse chairs and users through physics-driven synthetic data augmentation and rigorous cross-validation protocols.
Searching arXiv for ChairPose and closely related seated pose estimation / pose prior / posture correction papers. ChairPose is a wearable-free, vision-free system for full-body 3D seated pose estimation that uses pressure sensing together with explicit chair morphology to reconstruct a person’s pose while sitting on arbitrary chairs (Ray et al., 3 Aug 2025). It is designed to avoid the characteristic limitations of vision-based systems—occlusion, lighting dependence, viewpoint sensitivity, and privacy concerns—and of wearable systems, which are intrusive, uncomfortable for long-term use, and require precise placement and calibration (Ray et al., 3 Aug 2025). In the paper’s formulation, the distinctive technical claim is not merely that pressure sensing can infer seated posture, but that chair morphology must be incorporated explicitly because chair shape both alters pressure patterns and constrains feasible body configurations (Ray et al., 3 Aug 2025).
1. Problem setting and conceptual scope
ChairPose is situated in the broader problem of seated posture monitoring for ergonomics, healthcare, rehabilitation, and adaptive interfaces (Ray et al., 3 Aug 2025). The work begins from the observation that prolonged sitting is linked to musculoskeletal disorders, cardiovascular issues, discomfort, and reduced cognitive performance, while existing sensing paradigms impose substantial deployment tradeoffs (Ray et al., 3 Aug 2025). Vision-based methods can estimate rich 2D or 3D pose, but they are susceptible to occlusion and are often undesirable in homes, offices, and clinics for privacy reasons; wearable methods avoid cameras but remain intrusive and calibration-sensitive (Ray et al., 3 Aug 2025).
Within this landscape, prior pressure-based systems are described as attractive because they are passive and privacy-preserving, yet they are typically limited either to coarse posture classes or to flat or chair-specific surfaces (Ray et al., 3 Aug 2025). ChairPose’s contribution is therefore framed as a chair-agnostic seated pose estimator rather than a chair-specific classifier. The paper explicitly presents it as “the first full body, wearable free seated pose estimation system that relies solely on pressure sensing and operates independently of chair geometry” (Ray et al., 3 Aug 2025).
A common ambiguity concerns the meaning of “ChairPose.” In this literature, the term refers to a seated human-pose estimation system, not to the yoga posture “Chair Pose.” A related yoga paper describes a hybrid Edge-AI framework for real-time yoga pose detection and posture correction, but it does not discuss Chair Pose as a named, pose-specific case study (Haldar, 23 Mar 2026). This distinction matters because the seated-pose problem addressed by ChairPose is centered on pressure distributions and chair geometry rather than monocular pose estimation and asana classification.
2. Sensing substrate and chair-morphology grounding
The sensing hardware is a thin, flexible, chair-agnostic pressure sensing mattress rather than a sensorized chair (Ray et al., 3 Aug 2025). The reported platform is a Sensing.Tex Fitness Mat with an sensor grid, coverage area mm, sensor size mm, pressure range $0$–$5000$ mmHg, and non-slip, non-elastic thermoplastic polyolefin material (Ray et al., 3 Aug 2025). The mat is centered on the seat and, when possible, wrapped around the backrest to improve coverage (Ray et al., 3 Aug 2025). This design allows deployment across multiple chair types without embedding sensors into the furniture.
The paper’s central modeling move is to pair pressure with chair morphology. Chair geometry is acquired with a 3D scanner and represented during inference as a point cloud , which is encoded using a pre-trained PointNet-based feature extractor (Ray et al., 3 Aug 2025). This is not a peripheral feature; it is the mechanism by which the system reasons about how distinct seat and backrest geometries deform the pressure-to-pose mapping (Ray et al., 3 Aug 2025).
Ground-truth pose supervision is obtained from a TrueDepth camera-based 3D estimation pipeline (Ray et al., 3 Aug 2025). More specifically, the dataset uses Apple iPhone 16 Pro TrueDepth at about $2$ m distance and front-left view, with synchronization by a manual protocol in which participants tap the seat three times at the beginning and end of each session (Ray et al., 3 Aug 2025). This suggests that ChairPose is trained with visual supervision but intended to operate without vision at deployment.
The idea that seat geometry matters has precedents outside posture sensing. “Pose to Seat: Automated Design of Body-Supporting Surfaces” treats a specified sitting pose as the input to a computational pipeline that synthesizes a support surface while taking pressure distribution, frictional contact forces, and body torques into account (Leimer et al., 2020). A plausible implication is that ChairPose and pose-driven seating design attack complementary directions of the same coupling: one infers pose from support interaction, while the other designs support from pose.
3. Generative architecture: MotionQuantizer and Pressure2Pose
ChairPose uses a two-stage generative model (Ray et al., 3 Aug 2025). The first stage, MotionQuantizer (MQ), learns a discrete latent codebook for seated motion using a Vector Quantized Variational Autoencoder with a U-Net-based encoder-decoder architecture (Ray et al., 3 Aug 2025). The second stage, Pressure2Pose (P2P), is an autoregressive classifier that predicts tokenized pose sequences from pressure input, chair geometry, and the previous token, after which the pretrained MQ decoder reconstructs continuous 3D joint coordinates (Ray et al., 3 Aug 2025).
The MQ input is a motion segment spanning seconds with frames, 0 joints, and 1 dimensions (Ray et al., 3 Aug 2025). Pose is represented as SMPL parameters 2, converted to 3D joints 3 through forward kinematics, then augmented with linear velocity, angular velocity, linear acceleration, and angular acceleration (Ray et al., 3 Aug 2025). The resulting representation is
4
Training uses a weighted combination of reconstruction and quantization loss,
5
with annealed weights so that reconstruction is emphasized early and quantization later (Ray et al., 3 Aug 2025).
Two stabilization mechanisms are reported. First, the codebook is updated by exponential moving average,
6
with 7 (Ray et al., 3 Aug 2025). Second, quantization dropout randomly removes discrete tokens during training,
8
with 9 (Ray et al., 3 Aug 2025).
Pressure2Pose receives, at each timestep 0, pressure data 1, chair geometry 2, and the previous tokenized pose 3 (Ray et al., 3 Aug 2025). At 4, the previous token is replaced with a learned start-of-sequence vector 5 (Ray et al., 3 Aug 2025). The predicted token 6 is fed forward autoregressively, and the decoded output yields 7 (Ray et al., 3 Aug 2025).
The P2P objective combines token reconstruction with a sequence-level smoothness term: 8
9
$0$0
with $0$1 (Ray et al., 3 Aug 2025).
This tokenization strategy places ChairPose in methodological proximity to explicit pose-prior work, although the purposes differ. “QPoser” also compresses human poses into structured latent representations, but it is designed for controllable pose generation and editing rather than pressure-conditioned seated-pose inference (Li et al., 2023). More generally, work on pose priors argues that valid human pose lies on a correlated manifold not captured by simple per-joint limits or a single Gaussian, and that learned nonlinear priors are better suited to representing physical plausibility (Raaj, 2019). This suggests why discretized motion structure can be advantageous in a seated inference pipeline.
4. Simulation-assisted training and dataset design
A major challenge for ChairPose is limited real seated-motion data across diverse chairs and users. To address this, the paper introduces a physics-driven synthetic data pipeline adapted from PresSim for non-planar chair surfaces (Ray et al., 3 Aug 2025). The original floor-deformation module is replaced with 12 CAD chair models sourced from the internet, and RagdollXBlender is used to add ragdoll physics to motion capture files (Ray et al., 3 Aug 2025). The simulated body is composed of interconnected rigid parts with joints, collisions, gravity, and mass distribution, producing chair-specific body-seat interactions such as falling into a seat, adjusting balance, and settling posture (Ray et al., 3 Aug 2025).
The synthetic augmentation output contains 12 additional chair shapes and 1,153,500 frames with synchronized SMPL pose and pressure dynamics (Ray et al., 3 Aug 2025). This synthetic data is combined with the real dataset during training, while evaluation is performed only on real data (Ray et al., 3 Aug 2025). The paper reports that simple transformation-based augmentation—horizontal flipping, random scaling, and Gaussian noise—performed worse than both physics-driven augmentation and even real-data-only training (Ray et al., 3 Aug 2025).
The real dataset introduced by the work is TDSD (Temporal Dynamic Sitting Dataset) (Ray et al., 3 Aug 2025). It contains 8 participants, 4 chair types, 12 seated activities, and 96,125 synchronized real frames (Ray et al., 3 Aug 2025). The four chairs are office chair, foldable chair, barstool, and manual wheelchair (Ray et al., 3 Aug 2025). Each participant performed all 12 actions five times per chair, continuously and without breaks, so that action transitions were also captured (Ray et al., 3 Aug 2025).
The experimental setup uses PyTorch on an NVIDIA A6000 Ada Lovelace GPU and AMD Ryzen 9 CPU, with batch size 32, learning rate 1e-4, optimizer AdamW, weight decay 1e-5, maximum 200 epochs, early stopping patience 15, and a cosine annealing scheduler (Ray et al., 3 Aug 2025). Motion-quantization settings include codebook size $0$2, quantization window $0$3 second, EMA smoothing $0$4, quantization dropout $0$5, and sequence-loss weight $0$6 (Ray et al., 3 Aug 2025).
5. Evaluation protocol and quantitative performance
ChairPose is evaluated under three cross-validation regimes designed to measure generalization to unseen users, unseen chairs, and both jointly (Ray et al., 3 Aug 2025). These are LOUOCV (Leave-One-User-Out Cross-Validation), LOCOCV (Leave-One-Chair-Out Cross-Validation), and LOCUOCV (Leave-One-(User+Chair)-Out Cross-Validation) (Ray et al., 3 Aug 2025). The paper compares against prior pressure-based baselines IC, IS, and 3DHPE, as well as vision-based methods SMPLer-X and MediaPipe, and an internal direct-regression baseline that omits motion quantization and temporal modeling (Ray et al., 3 Aug 2025).
The best MQ configuration is reported as $0$7 second and $0$8, yielding $0$9, $5000$0, and R-Precision $5000$1 (Ray et al., 3 Aug 2025). Stepwise ablation shows consistent improvement from the feature extractor, EMA, and quantization dropout, with metrics progressing from MQ alone at 82.50 / 48.10 / 0.460 to MQ + FE + EMA + Q-D at 58.18 / 28.15 / 0.656 (Ray et al., 3 Aug 2025).
The main pose-estimation results are summarized below.
| Setting | MPJPE | PA-MPJPE | MPVE |
|---|---|---|---|
| LOUOCV | 53.5 mm | 37.9 mm | 44.2 mm |
| LOCOCV | 88.2 mm | 77.1 mm | 81.6 mm |
| LOCUOCV | 89.4 mm | 88.7 mm | 84.3 mm |
These figures correspond to the full ChairPose model (Ray et al., 3 Aug 2025). In the most difficult setting, where both user and chair are unseen, the reported mean per joint position error is 89.4 mm, the headline result also emphasized in the abstract (Ray et al., 3 Aug 2025).
Ablation against the internal baseline indicates that motion quantization and temporal modeling drive a large portion of the gain (Ray et al., 3 Aug 2025). In LOUOCV, for example, the direct baseline reports 195.2 / 169.7 / 181.4, while the full model reports 53.5 / 37.9 / 44.2 (Ray et al., 3 Aug 2025). Physics-based augmentation also materially improves robustness: MPJPE is reduced from 71.8 mm to 53.5 mm in LOUOCV, from 158.7 mm to 88.2 mm in LOCOCV, and from 164.2 mm to 89.4 mm in LOCUOCV (Ray et al., 3 Aug 2025). The text summarizes these as approximately 25%, 44%, and 46% relative reductions (Ray et al., 3 Aug 2025).
The comparison with camera methods is instructive rather than directly symmetric. MediaPipe reports 47.43 / 38.1 in LOUOCV, while SMPLer-X reports 11.2 / 10.1 / 11.38 (Ray et al., 3 Aug 2025). ChairPose does not exceed those vision numbers, but it is not intended to: it trades off some accuracy for privacy preservation, occlusion freedom, and wearable-free deployment (Ray et al., 3 Aug 2025).
6. Applications, limitations, and relation to adjacent research
The paper argues that ChairPose enlarges the design space for posture-aware systems by providing privacy-preserving sensing, no line-of-sight requirement, no wearable discomfort, robustness to occlusion and lighting, and compatibility with many chair types (Ray et al., 3 Aug 2025). In ergonomics, it is reported to detect slouching, asymmetry, spinal deviation, and related posture issues, with lumbar flexion error 4.2° and thoracic tilt error 3.8° (Ray et al., 3 Aug 2025). In healthcare, the recovered pose supports volumetric center-of-mass estimation: direct pressure-to-VCoM regression yields MAE 127.5 and RMSE 165.2, whereas the SMPL-pose-based approach yields MAE 31.4 and RMSE 42.8 (Ray et al., 3 Aug 2025). For seated action recognition, combining pressure and ChairPose-derived pose improves ALS-HAR Macro F1 from 0.813 to 0.824 and WS-HAR Macro F1 from 0.834 to 0.853 (Ray et al., 3 Aug 2025).
The work also sits alongside a broader transition from coarse sitting-posture recognition toward richer articulated inference. “LSP-YOLO” targets six sitting-posture categories with a lightweight single-stage embedded model and reports 94.2% accuracy and 251 Fps for its smallest PC configuration, but it addresses classification rather than full-body 3D pose reconstruction (Li et al., 18 Nov 2025). “WheelPoser,” by contrast, estimates upper-body pose for wheelchair users with four IMUs and reports 14.30° mean joint angle error and 6.74 cm mean joint position error, but it remains a wearable sparse-IMU system rather than a pressure-only chair-agnostic one (Li et al., 2024). ChairPose occupies a distinct point in this design space: full-body seated pose, no camera, no wearables, explicit seat-shape conditioning (Ray et al., 3 Aug 2025).
The paper is explicit about limitations. Deformable seat dynamics are not well modeled, so soft cushions can distort pressure patterns (Ray et al., 3 Aug 2025). Arm and head estimation are less accurate because those body parts often contribute weakly or indirectly to pressure (Ray et al., 3 Aug 2025). The autoregressive architecture can accumulate error over time and induce drift (Ray et al., 3 Aug 2025). There is also a responsiveness–accuracy tradeoff: a fast baseline mode has latency under about 0.11 s but higher error, while the full MQ + P2P mode is more accurate but has about 2 seconds startup latency because of the temporal window (Ray et al., 3 Aug 2025).
A plausible implication is that future seated-pose systems may become hybrid rather than purely pressure-based, especially for upper-body articulation. The paper itself suggests incorporating soft-material physics, collision-aware or physically constrained losses, additional complementary sensors such as IMUs for upper-body parts, broader user and chair diversity, and lower-latency variants (Ray et al., 3 Aug 2025). That trajectory would align with a general pattern in pose research: explicit priors and structured latent representations improve physical plausibility (Raaj, 2019, Li et al., 2023), while specialized sensing configurations become necessary when standard ambulatory assumptions fail (Li et al., 2024).
In sum, ChairPose defines a specific technical program for seated pose estimation: pressure-only sensing on a portable chair-agnostic mat, explicit conditioning on chair morphology, tokenized motion representation, autoregressive pressure-to-pose decoding, and physics-assisted augmentation for cross-chair and cross-user generalization (Ray et al., 3 Aug 2025). Its significance lies less in surpassing camera-based pose accuracy than in establishing that full-body seated pose can be inferred with no line of sight and no wearables, provided that seat geometry and motion structure are modeled jointly (Ray et al., 3 Aug 2025).
