pith. sign in

arxiv: 2604.06174 · v1 · submitted 2026-02-02 · 💻 cs.HC · cs.CY

X-BCD: Explainable Sensor-Based Behavioral Change Detection in Smart Home Environments

Gabriele Civitarese , Claudio Bettini This is my paper

Pith reviewed 2026-05-16 08:02 UTC · model grok-4.3

classification 💻 cs.HC cs.CY
keywords behavioral change detectionsmart home sensorsexplainable AImild cognitive impairmentactivity routineschange point detectionunsupervised learningsensor data analysis
0
0 comments X

The pith

X-BCD detects shifts in daily activity routines from smart home sensors and turns them into plain-language descriptions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Behavioral changes in home activities can mark cognitive decline, yet they stay hard to track through occasional clinic visits or raw sensor streams. X-BCD processes continuous multimodal sensor data with an unsupervised pipeline that locates change points and follows how activity clusters evolve. Detected reorganizations are then mapped to interpretable features and rendered as natural-language statements. A preliminary study on longitudinal records from real patients with mild cognitive impairment indicates the descriptions are consistent with cohort patterns and expert review.

Core claim

X-BCD is an explainable unsupervised framework that combines change point detection and cluster evolution tracking on multimodal smart home sensor data to identify and characterize changes in activity routines, then converts those changes into natural-language explanations grounded in interpretable features for clinical use.

What carries the argument

Unsupervised change point detection combined with cluster evolution tracking, which locates transitions in activity patterns and produces feature-based natural language summaries.

If this is right

  • Continuous home monitoring can supply clinicians with ongoing descriptions of routine changes instead of relying only on sporadic visits.
  • The generated explanations support decision support by highlighting simplifications or fragmentations in daily habits.
  • Parameter sensitivity checks allow the system to remain stable across different sensor setups and homes.
  • Cohort-level comparisons and expert review can validate that the outputs reflect real behavioral patterns.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Connecting the detected changes to specific cognitive test scores over time could turn them into stronger early indicators.
  • The same pipeline might extend to other conditions where shifts in home routines carry health meaning.
  • Integration with additional home devices could refine the granularity of the activity clusters being tracked.

Load-bearing premise

Changes found by unsupervised analysis of sensor streams correspond to clinically meaningful reorganizations of daily behavior without labeled examples or direct comparison to cognitive test results.

What would settle it

A longitudinal comparison in which X-BCD outputs are checked against repeated clinical cognitive assessments in the same patients to test whether the timing and type of detected changes align with measured functional decline.

Figures

Figures reproduced from arXiv: 2604.06174 by Claudio Bettini, Gabriele Civitarese.

Figure 1
Figure 1. Figure 1: The high-level pipeline of X-BCD. This pipeline is applied independently to each monitored behavioral dimension. activity levels, or nutrition-related patterns) and include explicit temporal context variables to account for circadian and weekly regularities. The resulting multivariate time series provides a structured representation of behavior over time. Next, the Change Point Detection module is in charg… view at source ↗
Figure 2
Figure 2. Figure 2: Graphical intuition behind evolutionary continuity and topological changes. [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Difference in number of changes across cohorts for each dimension [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Distribution of unstable routine mass (i.e., drift, split, merge, novel, and disappeared) across cohorts for each dimension [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Distribution of change types across cohorts [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of change types considering a single comprehensive sleep dimension [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Impact of parameter 𝛽 on the different dimensions across cohorts capture transient fluctuations, whereas very large values lead to over-smoothing and few or no detected changes. Between these regimes, a broad plateau emerges in which the number of detected change points remains stable across dimensions. The selected value 𝛽 = 60 lies within this plateau, indicating robustness to moderate parameter perturba… view at source ↗
Figure 8
Figure 8. Figure 8: Sleep stages: sensitivity analysis of drift and similarity threshold on cluster evolution tracking. [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
read the original abstract

Behavioral changes in daily life activities at home can be digital markers of cognitive decline. However, such changes are difficult to assess through sporadic clinical visits and remain challenging to interpret from continuous in-home sensing data. Extensive work has been done in the ubiquitous computing area on recognizing activities in smart homes, but only limited efforts have focused on analysing the evolution of patterns of activities, hence identifying behavior changes. In particular, understanding how daily habits and routines evolve and reorganize (e.g., simplification, fragmentation) is still an open challenge for clinical monitoring and decision support. In this paper, we present X-BCD, an explainable, unsupervised framework for detecting and characterizing changes in activity routines from multimodal smart home sensor data, combining change point detection and cluster evolution tracking. To support clinical interpretation, detected changes in routines are transformed into natural-language explanations grounded in interpretable features. Our preliminary evaluation on longitudinal data from real MCI patients shows that X-BCD produces interpretable descriptions of behavioral change, as supported by cohort-level comparisons, expert assessment, and parameter sensitivity analysis.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript presents X-BCD, an explainable unsupervised framework for detecting and characterizing changes in activity routines from multimodal smart home sensor data. It combines change-point detection with cluster evolution tracking and generates natural-language explanations grounded in interpretable features. The central claim is that a preliminary evaluation on longitudinal data from real MCI patients demonstrates that X-BCD produces interpretable descriptions of behavioral change, supported by cohort-level comparisons, expert assessment, and parameter sensitivity analysis.

Significance. If the evaluation claims hold under more rigorous validation, the work would address a meaningful gap in ubiquitous computing by moving beyond activity recognition to the detection and clinical interpretation of routine reorganization as digital markers of cognitive decline. The unsupervised, explainable design is a positive step toward deployable monitoring tools, though its impact is currently limited by the preliminary nature of the supporting evidence.

major comments (2)
  1. [Abstract] Abstract: the claim that the preliminary evaluation 'shows that X-BCD produces interpretable descriptions of behavioral change' rests entirely on unquantified expert assessment and cohort-level comparisons; no metrics, cohort size, data duration, sensor modalities, or specific result tables are supplied, which is load-bearing for the central clinical-utility assertion.
  2. [Evaluation] Evaluation section (implied by abstract): the framework applies standard change-point and clustering methods without labeled ground truth or correlation against independent cognitive measures (e.g., MMSE, CDR) or longitudinal clinical outcomes; expert judgment alone cannot establish that detected reorganizations are clinically meaningful rather than artifacts of the chosen thresholds.
minor comments (1)
  1. [Abstract] The abstract refers to 'parameter sensitivity analysis' but provides no details on the ranges tested or stability of the natural-language explanations; adding a brief table or figure would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below, clarifying the preliminary nature of the evaluation while committing to revisions that improve transparency without overstating the current evidence.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that the preliminary evaluation 'shows that X-BCD produces interpretable descriptions of behavioral change' rests entirely on unquantified expert assessment and cohort-level comparisons; no metrics, cohort size, data duration, sensor modalities, or specific result tables are supplied, which is load-bearing for the central clinical-utility assertion.

    Authors: We agree the abstract is too terse on evaluation details. In the revised manuscript we will expand it to report the cohort size, data collection duration, sensor modalities, and explicit references to the tables showing cohort-level comparisons and expert ratings. This will make the supporting evidence for the interpretability claim more concrete while retaining the 'preliminary' qualifier. revision: yes

  2. Referee: [Evaluation] Evaluation section (implied by abstract): the framework applies standard change-point and clustering methods without labeled ground truth or correlation against independent cognitive measures (e.g., MMSE, CDR) or longitudinal clinical outcomes; expert judgment alone cannot establish that detected reorganizations are clinically meaningful rather than artifacts of the chosen thresholds.

    Authors: The study is observational and unsupervised; no ground-truth labels or MMSE/CDR correlations were collected, so we cannot add such analyses. We will insert a dedicated limitations subsection that explicitly states these constraints, notes that expert assessment and sensitivity analysis provide only initial support, and outlines future validation plans against clinical outcomes. The current framing already presents the work as a preliminary framework rather than a clinically validated tool. revision: partial

Circularity Check

0 steps flagged

No significant circularity; framework applies standard unsupervised techniques with independent evaluation

full rationale

The paper describes X-BCD as an unsupervised combination of change-point detection and cluster evolution tracking on multimodal sensor data, followed by transformation to natural-language explanations. The central claim of producing interpretable behavioral change descriptions is supported by cohort comparisons, expert assessment, and parameter sensitivity analysis on real longitudinal MCI data. No equations, fitted parameters renamed as predictions, self-definitional loops, or load-bearing self-citations appear in the derivation. The approach relies on established methods without reducing outputs to inputs by construction, making the framework self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Abstract-only review; framework assumes standard change-point algorithms and clustering can capture routine evolution, with parameters tuned via sensitivity analysis.

free parameters (1)
  • change-point detection thresholds and clustering parameters
    Mentioned via parameter sensitivity analysis but values and fitting procedure not described in abstract.
axioms (1)
  • domain assumption Activity patterns extracted from multimodal sensors can be clustered and their temporal evolution tracked to reveal meaningful behavioral reorganizations.
    Core premise of the cluster evolution tracking component.

pith-pipeline@v0.9.0 · 5483 in / 1090 out tokens · 24914 ms · 2026-05-16T08:02:38.601067+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages

  1. [1]

    S., and Al Ayidh, A.Explainable ai for sensor signal interpretation to revolutionize human health monitoring: A review.IEEE Access(2025)

    Alharthi, A., Alqrashi, A., Alharbi, T., Alammar, M., Aldosari, N., Bouchekara, H., Shaaban, Y., Shahriar, M. S., and Al Ayidh, A.Explainable ai for sensor signal interpretation to revolutionize human health monitoring: A review.IEEE Access(2025)

    work page 2025

  2. [2]

    J.A survey of methods for time series change point detection.Knowledge and information systems 51, 2 (2017), 339–367

    Aminikhanghahi, S., and Cook, D. J.A survey of methods for time series change point detection.Knowledge and information systems 51, 2 (2017), 339–367

    work page 2017

  3. [3]

    J.Using change point detection to automate daily activity segmentation

    Aminikhanghahi, S., and Cook, D. J.Using change point detection to automate daily activity segmentation. In2017 IEEE international conference on pervasive computing and communications workshops (PerCom workshops)(2017), IEEE, pp. 262–267

    work page 2017

  4. [4]

    International Journal of Geriatric Psychiatry: A journal of the psychiatry of late life and allied sciences 25, 3 (2010), 224–233

    Aretouli, E., and Brandt, J.Everyday functioning in mild cognitive impairment and its relationship with executive cognition. International Journal of Geriatric Psychiatry: A journal of the psychiatry of late life and allied sciences 25, 3 (2010), 224–233

    work page 2010

  5. [5]

    [6]Association, A., et al.More than normal aging: understanding mild cognitive impairment.Alzheimers Dement

    Arrotta, L., Civitarese, G., Chen, X., Cumin, J., and Bettini, C.Multi-subject human activities: A survey of recognition and evaluation methods based on a formal framework.Expert Systems with Applications 267(2025), 126178. [6]Association, A., et al.More than normal aging: understanding mild cognitive impairment.Alzheimers Dement. 18(2022), 545–868

    work page 2025

  6. [6]

    Babangida, L., Perumal, T., Mustapha, N., and Yaakob, R.Internet of things (iot) based activity recognition strategies in smart homes: A review.IEEE sensors journal 22, 9 (2022), 8327–8336

    work page 2022

  7. [7]

    [9]Chen, C., Ding, S., and W ang, J.Digital health for aging populations.Nature medicine 29, 7 (2023), 1623–1630

    Bosc, M., Heitz, F., Armspach, J.-P., Namer, I., Gounot, D., and Rumbach, L.Automatic change detection in multimodal serial mri: application to multiple sclerosis lesion evolution.NeuroImage 20, 2 (2003), 643–656. [9]Chen, C., Ding, S., and W ang, J.Digital health for aging populations.Nature medicine 29, 7 (2023), 1623–1630

    work page 2003

  8. [8]

    Civitarese, G., Fiori, M., Arighi, A., Galimberti, D., Florio, G., and Bettini, C.The serenade project: Sensor-based explainable detection of cognitive decline.arXiv preprint arXiv:2504.08877(2025)

    work page arXiv 2025

  9. [9]

    N., Cook, D

    Dawadi, P. N., Cook, D. J., and Schmitter-Edgecombe, M.Automated cognitive health assessment using smart home monitoring of complex tasks.IEEE transactions on systems, man, and cybernetics: systems 43, 6 (2013), 1302–1313

    work page 2013

  10. [10]

    N., Cook, D

    Dawadi, P. N., Cook, D. J., and Schmitter-Edgecombe, M.Modeling patterns of activities using activity curves.Pervasive and mobile computing 28(2016), 51–68

    work page 2016

  11. [11]

    [14]Fahad, L

    Dwivedi, R., Dave, D., Naik, H., Singhal, S., Omer, R., Patel, P., Qian, B., Wen, Z., Shah, T., Morgan, G., et al.Explainable ai (xai): Core ideas, techniques, and solutions.ACM computing surveys 55, 9 (2023), 1–33. [14]Fahad, L. G., and Tahir, S. F.Activity recognition and anomaly detection in smart homes.Neurocomputing 423(2021), 362–372

    work page 2023

  12. [12]

    Fiori, M., Civitarese, G., Choudhary, P., and Bettini, C.Leveraging large language models for explainable activity recognition in smart homes: A critical evaluation.ACM Transactions on Internet of Things 6, 4 (2025), 1–25

    work page 2025

  13. [13]

    A., and Gargoum, A

    Gargoum, S. A., and Gargoum, A. S.Limiting mobility during covid-19, when and to what level? an international comparative study using change point analysis.Journal of Transport & Health 20(2021), 101019. X-BCD: Explainable Sensor-Based Behavioral Change Detection in Smart Home Environments•27

    work page 2021

  14. [14]

    C., Ritchie, K., Broich, K., Belleville, S., Brodaty, H., Bennett, D., Chertkow, H., et al.Mild cognitive impairment.The lancet 367, 9518 (2006), 1262–1270

    Gauthier, S., Reisberg, B., Zaudig, M., Petersen, R. C., Ritchie, K., Broich, K., Belleville, S., Brodaty, H., Bennett, D., Chertkow, H., et al.Mild cognitive impairment.The lancet 367, 9518 (2006), 1262–1270

    work page 2006

  15. [15]

    F., Preum, S

    Hoqe, E., Dickerson, R. F., Preum, S. M., Hanson, M., Barth, A., and Stankovic, J. A.Holmes: A comprehensive anomaly detection system for daily in-home activities. In2015 International Conference on Distributed Computing in Sensor Systems(2015), IEEE, pp. 40–51

    work page 2015

  16. [16]

    Johansson, M. M., Marcusson, J., and Wressle, E.Cognitive impairment and its consequences in everyday life: experiences of people with mild cognitive impairment or mild dementia and their relatives.International psychogeriatrics 27, 6 (2015), 949–958

    work page 2015

  17. [17]

    Khodabandehloo, E., Riboni, D., and Alimohammadi, A.Healthxai: Collaborative and explainable ai for supporting early diagnosis of cognitive decline.Future Generation Computer Systems 116(2021), 168–189

    work page 2021

  18. [18]

    A.Optimal detection of changepoints with a linear computational cost.Journal of the American Statistical Association 107, 500 (2012), 1590–1598

    Killick, R., Fearnhead, P., and Eckley, I. A.Optimal detection of changepoints with a linear computational cost.Journal of the American Statistical Association 107, 500 (2012), 1590–1598

    work page 2012

  19. [19]

    S., Bennett, D

    Li, P., Gao, L., Gaba, A., Yu, L., Cui, L., Fan, W., Lim, A. S., Bennett, D. A., Buchman, A. S., and Hu, K.Circadian disturbances in alzheimer’s disease progression: a prospective observational cohort study of community-based older adults.The Lancet Healthy Longevity 1, 3 (2020), e96–e105

    work page 2020

  20. [20]

    Liciotti, D., Bernardini, M., Romeo, L., and Frontoni, E.A sequential deep learning application for recognising human activities in smart homes.Neurocomputing 396(2020), 501–513

    work page 2020

  21. [21]

    Liguori, C., Placidi, F., Izzi, F., Spanetta, M., Mercuri, N. B., and Di Pucchio, A.Sleep dysregulation, memory impairment, and csf biomarkers during different levels of neurocognitive functioning in alzheimer’s disease course.Alzheimer’s research & therapy 12, 1 (2020), 5

    work page 2020

  22. [22]

    Neural Networks 43(2013), 72–83

    Liu, S., Y amada, M., Collier, N., and Sugiyama, M.Change-point detection in time-series data by relative density-ratio estimation. Neural Networks 43(2013), 72–83

    work page 2013

  23. [23]

    IEEE journal of biomedical and health informatics 23, 2 (2018), 838–847

    Lussier, M., Lavoie, M., Giroux, S., Consel, C., Guay, M., Macoir, J., Hudon, C., Lorrain, D., Talbot, L., Langlois, F., et al.Early detection of mild cognitive impairment with in-home monitoring sensor technologies using functional measures: a systematic review. IEEE journal of biomedical and health informatics 23, 2 (2018), 838–847. [27]Miller, T.Explan...

    work page 2018

  24. [24]

    Oliveira, M., and Gama, J.A framework to monitor clusters evolution applied to economy and finance problems.Intelligent Data Analysis 16, 1 (2012), 93–111

    work page 2012

  25. [25]

    Pérès, K., Helmer, C., Amieva, H., Orgogozo, J.-M., Rouch, I., Dartigues, J.-F., and Barberger-Gateau, P.Natural history of decline in instrumental activities of daily living performance over the 10 years preceding the clinical diagnosis of dementia: a prospective population-based study.Journal of the American Geriatrics Society 56, 1 (2008), 37–44

    work page 2008

  26. [26]

    Prenkaj, B., and Velardi, P.Unsupervised detection of behavioural drifts with dynamic clustering and trajectory analysis.IEEE Transactions on Knowledge and Data Engineering 36, 5 (2023), 2257–2270

    work page 2023

  27. [27]

    H., and Helaoui, R.Smartfaber: Recognizing fine-grained abnormal behaviors for early detection of mild cognitive impairment.Artificial intelligence in medicine 67(2016), 57–74

    Riboni, D., Bettini, C., Civitarese, G., Janjua, Z. H., and Helaoui, R.Smartfaber: Recognizing fine-grained abnormal behaviors for early detection of mild cognitive impairment.Artificial intelligence in medicine 67(2016), 57–74

    work page 2016

  28. [28]

    Rozzini, L., Chilovi, B. V., Peli, M., Conti, M., Rozzini, R., Trabucchi, M., and Padovani, A.Anxiety symptoms in mild cognitive impairment.International Journal of Geriatric Psychiatry: A journal of the psychiatry of late life and allied sciences 24, 3 (2009), 300–305. [33]Selkoe, D. J.Alzheimer’s disease is a synaptic failure.Science 298, 5594 (2002), 789–791

    work page 2009

  29. [29]

    Song, D., Zhou, J., Ma, J., Chang, J., Qiu, Y., Zhuang, Z., Xiao, H., and Zeng, L.Sleep disturbance mediates the relationship between depressive symptoms and cognitive function in older adults with mild cognitive impairment.Geriatric Nursing 42, 5 (2021), 1019–1023

    work page 2021

  30. [30]

    InProceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining(2006), pp

    Spiliopoulou, M., Ntoutsi, I., Theodoridis, Y., and Schult, R.Monic: modeling and monitoring cluster transitions. InProceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining(2006), pp. 706–711

    work page 2006

  31. [31]

    Sprint, G., Cook, D. J., and Fritz, R.Behavioral differences between subject groups identified using smart homes and change point detection.IEEE journal of biomedical and health informatics 25, 2 (2020), 559–567

    work page 2020

  32. [32]

    Staudacher, M., Telser, S., Amann, A., Hinterhuber, H., and Ritsch-Marte, M.A new method for change-point detection developed for on-line analysis of the heart beat variability during sleep.Physica A: Statistical Mechanics and its Applications 349, 3-4 (2005), 582–596

    work page 2005

  33. [33]

    C.-I., Rawtaer, I., et al.Predicting mild cognitive impairment through ambient sensing and artificial intelligence

    Tan, A.-H., Ying, W.-Y., Subagdja, B., Huang, A., Tay, T. C.-I., Rawtaer, I., et al.Predicting mild cognitive impairment through ambient sensing and artificial intelligence. In2024 IEEE Conference on Artificial Intelligence (CAI)(2024), IEEE, pp. 1098–1104

    work page 2024

  34. [34]

    Expert Systems with Applications 205(2022), 117538

    Teh, S.-K., Rawtaer, I., and Tan, A.-H.Predictive self-organizing neural networks for in-home detection of mild cognitive impairment. Expert Systems with Applications 205(2022), 117538

    work page 2022

  35. [35]

    J., Blackwell, T., Stone, K

    Tranah, G. J., Blackwell, T., Stone, K. L., Ancoli-Israel, S., Paudel, M. L., Ensrud, K. E., Cauley, J. A., Redline, S., Hillier, T. A., Cummings, S. R., et al.Circadian activity rhythms and risk of incident dementia and mild cognitive impairment in older women. Annals of neurology 70, 5 (2011), 722–732

    work page 2011

  36. [36]

    apathy and depression in mild cognitive impairment: Distinct longitudinal trajectories and clinical outcomes

    Velayudhan, L.Apathy and depression as risk factors for dementia conversion in mild cognitive impairment: Commentary on “apathy and depression in mild cognitive impairment: Distinct longitudinal trajectories and clinical outcomes” by connors et al.International Psychogeriatrics 35, 11 (2023), 598–600

    work page 2023

  37. [37]

    I., Pan, S., Aggarwal, C., and Salehi, M.Deep learning for time series anomaly detection: A 28•Civitarese and Bettini survey.ACM Computing Surveys 57, 1 (2024), 1–42

    Zamanzadeh Darban, Z., Webb, G. I., Pan, S., Aggarwal, C., and Salehi, M.Deep learning for time series anomaly detection: A 28•Civitarese and Bettini survey.ACM Computing Surveys 57, 1 (2024), 1–42. Received 20 February 2007; revised 12 March 2009; accepted 5 June 2009

    work page 2024