{"id":"8aae7308-4764-4647-b31b-02c78ee0adc7","arxiv_id":"2411.16097","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A barometer-based wearable system with automatic KVLU calibration estimates load vertical location during lifting with 5.71 cm mean absolute error.","lead":"This paper proposes a wearable system of smart insoles and wristbands that estimates how high a worker holds a load during lifting, using air pressure instead of inertial sensors to avoid drift. It reports a mean error of 5.71 cm, which could improve workplace lifting-risk assessment if the auto-calibration holds in independent tests.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 5.71 cm MAE is not yet an independent accuracy estimate: Eq. 4–5 leave a and b undisclosed and the KVLU reference ratio is computed from the same subjects whose lifts are scored, so the headline may be an in-sample fit artifact.","rationale":"The paper has a concrete, falsifiable central claim and uses optical motion capture as ground truth, which is appropriate. The 5.71 cm MAE in Table IV is the number that carries the paper, and for that number to be an accuracy estimate rather than a fit residual, every quantity in Eq. 5 must be fixed independently of the scored trials. The weakest link is the calibration chain: LVL_KVLU is body height times 0.495, with 0.495 computed from the same ten subjects whose lifting trials are scored, and the linear parameters a and b are never reported. It is possible that a and b come from the earlier pilot study [16] and that the ratio is only a cohort mean; in that case the issue is a serious reporting gap that a leave-one-subject-out reanalysis would likely resolve. If, however, a or b were adjusted to make the lifting trials match, the 5.71 cm value is not an independent accuracy. The experimental protocol also only intersperses standing calibration between short, discrete lifts, so the longer-term and walking-based auto-calibration claims are not exercised by the LVL accuracy test. These considerations do not justify rejection: the hardware is unobtrusive, the KVLU idea is sensible, and the reference-point detection results in Tables I–III are informative. They do justify keeping the verdict conditional until the calibration parameters are disclosed and an out-of-sample evaluation is run.","tokens_in":13700,"tokens_out":8738,"duration_ms":84921,"concrete_test":"Request the authors to disclose a and b and the exact fitting protocol for Eq. 4, then recompute Table IV under leave-one-subject-out calibration: fit the ratio (and any a,b) on nine subjects and evaluate the held-out subject, rotating over all ten subjects. If the held-out MAE is materially worse than the reported 5.71 cm (e.g., exceeds 10 cm), the headline accuracy is an in-sample calibration artifact rather than an independent estimate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim rests on Eq. 5, LVL = a*(P − P_KVLU) + LVL_KVLU + b, but the paper never reports a, b, or how they were determined. If these parameters (or the 0.495 wrist-height/body-height ratio in Sec. IV.A.3, obtained from the same ten subjects whose later lifts appear in Table IV) were fit or tuned on the evaluation trials, the 5.71 cm MAE is an in-sample statistic rather than an independent accuracy. Even without intentional fitting, using the in-sample mean ratio makes the 2.04 cm standing-reference error of Table III optimistic for new users, and no leave-one-subject-out or external-cohort analysis is provided. In addition, the LVL accuracy experiment in Sec. IV.B uses only standing KVLU updates between 30 discrete lifts with 5 seconds of quiet standing; walking-based KVLU and long-duration drift are never tested, so the paper's long-term auto-calibration claim is not supported by the 5.71 cm number.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an unobtrusive wearable system (two smart insoles and two smart wristbands) that uses barometric pressure, IMU, and plantar pressure data to estimate the vertical location of a load (LVL) during manual lifting. To correct barometer drift, the authors introduce a Known Vertical Location Update (KVLU) calibration method that identifies reference postures during standing and walking, estimates the wrist height at those postures using an anthropometric ratio, and updates the pressure-to-height transformation. The system is evaluated on ten subjects, with a reported overall mean absolute error of 5.71 cm for LVL measurement, compared with 25 cm and 33 cm errors for IMU-based baselines. The abstract and conclusion claim that the KVLU method enables long-term, auto-calibrated LVL measurement in uncontrolled workplaces.","tokens_in":13987,"tokens_out":3676,"duration_ms":34430,"significance":"If the reported accuracy is independently validated, the system could be a practical, low-cost alternative to vision-based or multi-IMU motion capture for ergonomic risk assessment, particularly for NIOSH lifting-equation inputs. The contribution of an auto-calibration mechanism for barometer-based vertical tracking is conceptually appealing and relevant to wearable sensing for occupational health. The paper includes a hardware prototype, a clear experimental protocol with optical motion capture as ground truth, and quantitative comparisons to prior work. However, the central accuracy claim currently rests on an evaluation protocol that includes in-sample calibration parameters and undisclosed model coefficients, so the magnitude of the real-world benefit is not yet established.","major_comments":[{"comment":"The parameters a and b in the linear pressure-to-height model are never specified, and the text does not describe how they are determined. The central accuracy claim (MAE 5.71 cm) depends entirely on this transformation, but the reader cannot tell whether a and b are fixed physical constants (e.g., derived from the barometric formula) or fitted parameters, and if fitted, whether they were obtained from the same evaluation data. Please report the values, the calibration procedure, and any validation separate from the lifting trials.","section":"III.B.4, Eqs. (4)-(5)"},{"comment":"The wrist-height-to-body-height ratio (0.495) used as the 'known' wrist vertical location at KVLU reference points is computed from the same ten subjects whose lifting data are later scored in Table IV. This makes the 2.04 cm standing-reference error and the subsequent 5.71 cm LVL MAE partially in-sample statistics: the calibration constant is tuned to the evaluation population. Please provide a leave-one-subject-out analysis or a separate calibration cohort to demonstrate that the reported accuracy holds for unseen users.","section":"IV.A.3 and Table III"},{"comment":"The LVL accuracy experiment in Section IV.B uses only standing-based KVLU updates between 30 discrete lifts with 5 seconds of quiet standing; walking-based KVLU and long-duration drift (e.g., the 8-hour workplace scenario mentioned in the introduction) are not evaluated. The conclusion's claim that the KVLU method 'can effectively remove drift errors in LVL measurement' over long periods is therefore not supported by the reported experiment. Either add an experiment with walking-based updates and longer continuous monitoring, or restrict the claims to the tested scenario.","section":"IV.B and VI"}],"minor_comments":[{"comment":"The overall mean in the bottom-right cell appears to be 5.72 cm rather than 5.71 cm if computed from the rounded column means; please double-check the rounding convention and report the computation to one decimal place consistently.","section":"Table IV"},{"comment":"The sentence describing the ratio calculation should specify the posture (standing with wrists vertical) and clarify that 0.495 is a sample mean from ten subjects, not a general anthropometric constant from the literature.","section":"IV.A.3"},{"comment":"The notation in Equation (1) is inconsistent with the surrounding text ('nang l e' vs. 'n_angle' and 'T hr eshol d angle' vs. 'Threshold_angle'); please use a consistent symbol for the threshold.","section":"III.B.1, Eq. (1)"},{"comment":"The introduction states that barometers 'do not have integral drifting errors' while later acknowledging environment-induced drift; this apparent contradiction should be resolved by distinguishing integration drift from environmental pressure drift.","section":"I and III.A"},{"comment":"The text says the insole PCB is 'easy to be fixed on the shoelaces,' which seems inconsistent with an insole-embedded device; please verify this wording.","section":"III.A"}],"recommendation":"major_revision","confidential_remarks":"The paper has a promising concept and a usable hardware prototype, but the current evaluation does not yet support the headline accuracy because the calibration ratio is in-sample and the linear model coefficients are undisclosed. The missing long-duration and walking-based calibration experiments also weaken the main contribution. These are fixable with additional analysis or a revised experimental protocol, so the appropriate decision is major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the paper. The worthwhile thing here is the KVLU idea: borrowing ZUPT's known-update trick to correct barometric drift in wrist height. That's genuinely new in the wearable ergonomics space, and the hardware-plus-algorithm package is a real step beyond the 17-IMU and 5-IMU baselines they compare against. They built the insoles and wristbands, ran ten subjects through standing and walking trials, and used optical motion capture as ground truth. The walking reference-point detection rates in Table I and the 1.5-4.4 cm errors in Table II are useful, honest numbers. The wrist-vs-load proxy analysis in Tables VI and VII is also more open than most such papers.\n\nThe soft spots are real and they land on the headline number. The 5.71 cm MAE is not an independent accuracy estimate. The 'known' wrist height ratio (0.495) is computed from the same ten subjects whose lifts are then scored, so the standing calibration error of 2.04 cm and the LVL MAE are in-sample. It's not outright circularity in a regression sense, but it's optimistic for new users, and the fix is a leave-one-subject-out or held-out cohort analysis. More importantly, the linear mapping parameters a and b in Eq. 4-5 are never specified. Without them, nobody can reproduce the 5.71 cm or check whether they were tuned on the evaluation trials. That's a reproducibility gap, not a niche complaint.\n\nAlso, the LVL accuracy experiment only used standing KVLU updates, not the walking-based ones that are the paper's long-term selling point. So the 'reliable over 8 hours' claim is not backed by the reported test. The experiment covers four discrete lift heights with 30 lifts each, which is fine for a first validation, but there are no confidence intervals or per-condition variance aside from the MAE values.\n\nOverall: this is a promising systems paper with a real contribution, and it deserves a serious referee. But the central accuracy claim needs a revision that discloses the fitting procedure and parameters, and adds an independent validation. I'd send it out rather than desk-reject, and I'd ask for those specifics.","headline":"Genuinely new auto-calibration idea and plausible hardware, but the 5.71 cm MAE is an in-sample number until the calibration ratio and linear parameters are independently validated.","tokens_in":14492,"tokens_out":2680,"would_cite":false,"duration_ms":25081,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Wristband-and-insole barometers estimate load height during lifting to within about 5.7 cm, roughly five times more accurately than the IMU-based wearables they are compared against.","keywords":["wearable sensors","barometer","load vertical location","manual lifting","low-back pain","auto-calibration","gait analysis","inertial measurement unit"],"falsifier":"Recompute the wrist-height ratio (and the pressure-to-height slope $a$ and intercept $b$) on a separate, held-out group of subjects, then apply the pre-registered values to a new set of lifting trials with optical motion capture as ground truth. If the resulting MAE climbs substantially above 5.71 cm (e.g., beyond, say, 8 cm), the auto-calibration's accuracy claim is not transportable. A cheaper version: run the system over an 8-hour day with the same subjects, with all calibration points disabled for the first hour, and measure how quickly drift re-accumulates; KVLU claims to remove drift errors over long periods, so error should stay near 5.71 cm throughout.","tokens_in":13528,"feed_emoji":"📏","tokens_out":10715,"duration_ms":84438,"temperature":0.7,"pith_summary":"The paper claims that load vertical location (LVL) during manual lifting can be measured to a mean absolute error of 5.71 cm using two barometer-equipped wristbands and two pressure-sensing insoles, without the integration drift that limits IMU-based wearables. The central difficulty it addresses is barometric drift caused by changing environmental conditions. To fix that, it introduces a Known Vertical Location Update (KVLU) calibration: every time the system recognizes a standing or walking posture in which the wrist's height is predictable from body height, it resets the pressure-to-height relationship. If accurate, this gives workplace ergonomists a practical, unobtrusive tool for assessing low-back-pain risk, with error corresponding to about 1.7 percent in the lifting index rather than the 7.6–10 percent from earlier IMU approaches.","feed_headline":"Wearable barometer system tracks lift height to 5.7 cm","feed_subtitle":"Two wristbands and two insoles auto-calibrate during a normal workday, cutting lifting-risk scoring error to roughly 1.7%.","key_machinery":"The mechanism is the barometer-based vertical-location estimator combined with the Known Vertical Location Update (KVLU) auto-calibration. The wristband barometer measures air pressure, and vertical displacement is recovered from pressure change via the linear relation $\\mathrm{LVL} = a (P - P_{\\mathrm{KVLU}}) + \\mathrm{LVL}_{\\mathrm{KVLU}} + b$. KVLU supplies the reference: it recognizes postures where the wrist height is estimable from body height — standing with arms vertical, and the foot-flat phase of each gait cycle during walking — by fusing insole pressure signals (which identify standing and foot flat) with wrist IMU pitch angles (which identify the vertical wrist posture). At each detected reference point, the system updates $P_{\\mathrm{KVLU}}$ and the known wrist height $\\mathrm{LVL}_{\\mathrm{KVLU}}$, zeroing out drift errors that accumulate between calibrations. Anthropometry provides the known height: the paper computes an average wrist-height-to-body-height ratio of 0.495 from ten subjects and uses it to estimate the reference height.","core_discovery":"The central claim, stated in the abstract and conclusion, is that the proposed wearable system achieves a mean absolute error of 5.71 cm in LVL measurement during lifting, with errors of 4.31 cm at ground level, 4.76 cm at knee level, 6.60 cm at waist level, and 7.22 cm at shoulder level. The authors attribute this accuracy to replacing IMU-based integration with barometric pressure as the vertical-location signal and to the KVLU method, which removes environmental drift by repeatedly updating the reference pressure $P_{\\mathrm{KVLU}}$ whenever a known wrist vertical location is detected during ordinary standing and walking. They report that KVLU reference points are identified in nearly 100% of walking steps for most subjects, and that the estimated wrist height at these points has a mean absolute error between 1.47 cm and 4.38 cm across gait conditions. The authors also state that this is, to their knowledge, the only wearable LVL system with auto-calibration, and that its 5.71 cm error translates to roughly 1.73% error in lifting-risk assessment under the standard lifting equation, versus 10% and 7.6% for the 5-sensor and 17-sensor IMU baselines.","pith_inferences":["A direct test of transportability would be to recompute the 0.495 wrist-to-body-height ratio on an independent subject group, fix the pressure-to-height coefficients in advance, and then measure MAE on a fresh set of lifting trials.","The KVLU principle may generalize to other barometric wearables — for instance, fall detection or stair-climbing monitors — wherever a posture with a predictable body-segment height recurs often, though the paper does not explore those applications.","Because the system tracks the wrist rather than the load itself, field users may need a posture-dependent correction term; the paper's own tables show wrist-to-load offsets up to about 6 cm at ground level.","An 8-hour workplace trial with normal standing/walking cadence would reveal how quickly barometric drift returns between KVLU updates and whether the claimed long-term removal of drift errors holds outside a short laboratory session."],"forward_implications":["The 5.71 cm MAE translates to roughly a 1.73% error in the lifting index of the standard occupational lifting equation, versus 7.6–10% for the IMU baselines the paper compares against.","Nearly every walking step yields a KVLU reference point for at least one wrist, so calibration can be continuous throughout a work shift without interrupting the worker.","Opposite-side foot-and-wrist combinations (left foot with right wrist, and vice versa) produce both more frequent and more accurate KVLU reference points, giving a concrete design rule for sensor data fusion.","The observed systematic bias — wrist above the load at low lift heights and below the load at shoulder height — could be modeled and subtracted to lower the error further, as the paper itself notes as future work."],"supporting_citations":[{"why":"The standard manual that defines LVL and the sensitivity of lifting-risk scoring to LVL error, motivating the accuracy target.","marker":"[8]"},{"why":"The five-IMU wearable baseline that reported a 33 cm mean error, which the proposed system claims to surpass.","marker":"[14]"},{"why":"The seventeen-IMU wearable baseline that reported a 25 cm mean error, illustrating IMU drift over time.","marker":"[15]"},{"why":"The authors' prior pilot study using a wrist-worn barometer to track vertical travel during lifting, the foundation of the barometric method.","marker":"[16]"},{"why":"A characterization of barometric altimeter drift over time, the core problem KVLU is designed to correct.","marker":"[19]"},{"why":"A smart-insole gait recognition method used here to detect standing and foot-flat phases.","marker":"[27]"},{"why":"An anthropometric study connecting limb segment length to body height, the basis for estimating wrist height from body height.","marker":"[29]"},{"why":"An adaptive-threshold ZUPT algorithm, the inspiration for KVLU's known-value update strategy.","marker":"[30]"}],"fun_headline_variants":["Wearable barometer system auto-calibrates for lift height accuracy","Lift tracking to 5.7 cm with self-calibrating wearables","Barometer insoles and wristbands measure lift height to 5.7 cm","Auto-calibrating wearables cut lifting risk error to 1.7%","Smart insoles and bands track lift height with 5.7 cm error"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"To compute a 'known' wrist height during standing and walking, the paper uses a wrist-to-body-height ratio of 0.495 obtained from the same ten subjects whose later lifting trials are scored, and the linear pressure-to-height coefficients are never reported; if those values are adjusted to the evaluation data, the claimed 5.71 cm error would partly reflect curve fitting rather than genuine system accuracy.","fun_headline_variants_meta":{"raw":{"variants":["Wearable barometer system auto-calibrates for lift height accuracy","Lift tracking to 5.7 cm with self-calibrating wearables","Barometer insoles and wristbands measure lift height to 5.7 cm","Auto-calibrating wearables cut lifting risk error to 1.7%","Smart insoles and bands track lift height with 5.7 cm error"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000975,"raw_usage":{"total_tokens":4191,"prompt_tokens":1042,"completion_tokens":3149,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":658,"completion_tokens_details":{"reasoning_tokens":3046}},"tokens_in":658,"tokens_out":3149,"duration_ms":20528,"temperature":1.0,"reasoning_tokens":3046,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:32:30.334862+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the wrist-height ratio (and the pressure-to-height slope $a$ and intercept $b$) on a separate, held-out group of subjects, then apply the pre-registered values to a new set of lifting trials with optical motion capture as ground truth. If the resulting MAE climbs substantially above 5.71 cm (e.g., beyond, say, 8 cm), the auto-calibration's accuracy claim is not transportable. A cheaper version: run the system over an 8-hour day with the same subjects, with all calibration points disabled for the first hour, and measure how quickly drift re-accumulates; KVLU claims to remove drift errors over long periods, so error should stay near 5.71 cm throughout.","supporting_citations":[{"cited_title":"Applications manual for the revised niosh lifting equation,","cited_arxiv_id":null,"evidence_quote":"The standard manual that defines LVL and the sensitivity of lifting-risk scoring to LVL error, motivating the accuracy target."},{"cited_title":"Accuracy of estimating hand location during lifting using five wearable motion sensors,","cited_arxiv_id":null,"evidence_quote":"The five-IMU wearable baseline that reported a 33 cm mean error, which the proposed system claims to surpass."},{"cited_title":"Characterizing human box-lifting behav- ior using wearable inertial motion sensors,","cited_arxiv_id":null,"evidence_quote":"The seventeen-IMU wearable baseline that reported a 25 cm mean error, illustrating IMU drift over time."},{"cited_title":"The potential of wrist-worn barometer for measuring load vertical location in manual material lifting tasks: A pilot study,","cited_arxiv_id":null,"evidence_quote":"The authors' prior pilot study using a wrist-worn barometer to track vertical travel during lifting, the foundation of the barometric method."},{"cited_title":"The mems-based baromet- ric altimeter inaccuracy and drift phenomenon,","cited_arxiv_id":null,"evidence_quote":"A characterization of barometric altimeter drift over time, the core problem KVLU is designed to correct."},{"cited_title":"Bring gait lab to everyday life: Gait analysis in terms of activities of daily living,","cited_arxiv_id":null,"evidence_quote":"A smart-insole gait recognition method used here to detect standing and foot-flat phases."},{"cited_title":"Correlation of the stature to forearm length in the young adults of western indian population,","cited_arxiv_id":null,"evidence_quote":"An anthropometric study connecting limb segment length to body height, the basis for estimating wrist height from body height."},{"cited_title":"Adaptive threshold based zupt for single imu enabled wearable pedestrian localization,","cited_arxiv_id":null,"evidence_quote":"An adaptive-threshold ZUPT algorithm, the inspiration for KVLU's known-value update strategy."}],"review_version":1}