The researchers’ central insight is that human behavior is deeply individual. A generalized forecasting model, trained on the pooled habits of many people, may learn that most people make breakfast after waking up, but it will fail when a particular patient’s morning ritual diverges from the statistical average. For dependent individuals, such as older adults living alone or people managing chronic illness, those divergences are not noise to be smoothed away. They are the earliest signals that something has changed, potentially something alarming. The team therefore proposes an N-of-1 framework, a term borrowed from personalized medicine meaning a model built for a single subject rather than a population, that learns the behavioral fingerprint of one specific person from the life-logs generated by activity recognition modules.

Technically, the framework is organized as a hierarchy that captures human behavior at multiple levels of granularity simultaneously. At the lowest level, the system recognizes atomic actions, the short, indivisible units of movement such as reaching for a cup or opening a cabinet. At a higher level, it recognizes composite actions, which are sequences of atomic actions that together constitute meaningful activities such as preparing a meal or tidying a room. Existing systems, the authors note, typically categorize actions at only one of these levels, either unit actions, composite actions, or broad behaviors. By contrast, the new framework brings all levels together in a single pipeline, so that the model can reason about a person’s day both in fine-grained motion terms and in higher-level semantic terms.

The temporal backbone of the framework is the transformer, the attention-based architecture that has reshaped modern machine learning since its introduction in 2017. In this system, the transformer serves as an encoder of the temporal timeline: it takes the stream of recognized actions produced by the activity recognition modules and learns the patterns embedded in that sequence, capturing which actions tend to follow which, at what times, and in what contexts. Once those temporal patterns have been encoded, a decoder processes the resulting feature representations to forecast the next action in the sequence. This encode-then-decode design mirrors the way transformers handle language, but here the vocabulary consists of human actions rather than words, and the sentences are the unfolding routines of daily life.

What elevates the framework beyond simple next-action prediction is its use of the learned patterns for safety monitoring. Because the model builds a personalized picture of how a specific individual normally behaves, deviations from that picture become detectable. The learned patterns are used to recognize anomalies and to forecast alarming situations earlier than would otherwise be possible. In an assistive living context, this matters enormously: a system that only reacts after a fall or a medical emergency has occurred offers limited protection, whereas one that notices a person’s routine quietly drifting away from its usual shape can raise an alert while intervention is still easy and effective.

To validate the approach, the researchers conducted a case study focused on vision modalities, testing the framework on the Toyota SmartHome Untrimmed dataset, known as TSU. This dataset is an important benchmark because, unlike curated clips that show one clean action at a time, it consists of real-world, untrimmed video recordings of daily activities in smart home environments. Untrimmed footage is far closer to what an actual assistive system would face: long stretches of continuous living in which relevant actions are embedded in a stream of ordinary, unlabeled moments. The experimental results showed that the proposed framework achieves better accuracy compared to existing activity forecasting models on this challenging benchmark.

The study situates itself within a rich body of prior work on human activity recognition and prediction. Earlier research has explored activity recognition using wearable accelerometers, Wi-Fi channel state information, smartwatches paired with Bluetooth beacons, and depth cameras combined with inertial sensors. Other lines of work have applied long short-term memory networks to predict activities in smart homes, used change-point detection to segment activities more effectively, and built ontology-based systems to infer high-level context from low-level sensor data. Anomaly detection in the daily routines of older persons in single-resident smart homes has also been studied as a proof of concept. The new framework distinguishes itself by combining personalization, multi-level action categorization, and forecasting within one architecture, rather than treating these as separate problems.

The choice of the TSU dataset also reflects a deliberate methodological stance. The authors report that the data supporting their findings are available on reasonable request, and the research was approved by the research committee of the University of Engineering and Technology Lahore. Because the study relies on a publicly available dataset, the consent protocols followed for the subjects participating in the underlying data collection apply. The authors declare that no funds, grants, or other support were received during the preparation of the manuscript, and they report no relevant financial or non-financial conflicts of interest. All authors contributed to the study’s conception and design, with data collection, analysis, and experiments performed by Wahla, who also wrote the first stable draft.

The implications for assistive living are considerable. As populations age across much of the world, the demand for technologies that allow people to live independently and safely at home is growing faster than the supply of human caregivers. Systems like the one described in this study point toward a model of care in which the home itself becomes a quiet, continuous observer, one that knows its resident well enough to distinguish a harmless quirk from a warning sign. Personalization is the key to that distinction: a model that has learned one person’s rhythms can be sensitive to that person’s changes in a way no population-level model can.

There remain, of course, practical hurdles between a validated research framework and a deployed assistive product. The current validation rests on vision modalities, and real deployments would need to handle variable lighting, occlusion, privacy concerns around cameras, and the integration of other sensor types such as wearables and ambient sensors. The N-of-1 approach also implies that each new user requires a period during which the system learns their individual patterns before it can forecast and detect anomalies reliably. Nevertheless, the study demonstrates that treating behavior prediction as a personalized, multi-level, sequence-modeling problem, rather than a generalized classification problem, produces measurably better forecasting accuracy on realistic smart home data. For a field whose ultimate goal is to protect vulnerable people in their own homes, that shift in perspective, from the average person to the individual, may prove to be the most consequential contribution of all.

Subject of Research: Personalized transformer-based human activity forecasting for assistive living using smart home sensor data

Article Title: N-of-1 multi-modal human activity forecasting framework for assistive living

Article References: Wahla, S. Q., & Khan, M. U. G. (2026). N-of-1 multi-modal human activity forecasting framework for assistive living. Neural Computing and Applications, 38(19), Article 762. https://doi.org/10.1007/s00521-026-12079-y

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12079-y

Keywords: human activity forecasting, assistive living, smart homes, transformer, N-of-1 personalization, activity recognition, anomaly detection, Toyota SmartHome Untrimmed dataset, elderly care, machine learning, computer vision, behavioral modeling

The study, led by Johanna Hänggi of the Swiss Tropical and Public Health Institute and the University of Basel, set out to answer a deceptively simple question: does the way parents divide their waking hours between sitting, moving lightly, and exercising vigorously predict how their children divide the same hours? Previous evidence on this point has been inconclusive. Systematic reviews of children aged six to twelve support the idea that parents model activity for their offspring, particularly in father-son and mother-daughter pairs, but reviews restricted to device-based measurements found only weak positive correlations between parent and child activity. Longitudinal studies, which can reveal whether parental behavior actually shapes children’s behavior over time rather than merely coinciding with it, have been rare, and those that exist often relied on self-reported activity, which is vulnerable to recall errors and the human tendency to overstate virtuous behavior.

The SOPHYA team addressed these gaps with an unusually rigorous design. All children aged six to sixteen born and registered in Switzerland between 1998 and 2007 were eligible, and a random sample was drawn from the national population registry, stratified by birth year to ensure balanced age groups. Baseline assessments took place in 2013 to 2015 and follow-up assessments in 2019 to 2020. Rather than asking families to estimate their own activity, the researchers mailed pre-programmed ActiGraph accelerometers to participants, who wore the devices on the right hip for seven consecutive days while keeping diaries of non-wear periods. To be included, participants needed at least three weekdays of ten hours or more of wear time and at least one weekend day of eight hours, with weekday and weekend data weighted five to two to reflect a full week.

The baseline sample ultimately included 686 mother-child pairs and 373 father-child pairs with valid accelerometer and questionnaire data from the same measurement week, and 337 of those children provided valid follow-up data, yielding 263 mother-child and 149 father-child pairs for the longitudinal analysis. Sedentary behavior was defined as fewer than 100 counts per minute, while moderate-to-vigorous activity thresholds were age-adjusted for children and set at a level corresponding to four metabolic equivalents. Light activity occupied the range in between. From these raw counts the researchers computed each person’s waking movement composition: the proportions of wear time spent sedentary, in light activity, and in moderate-to-vigorous activity.

Analyzing those proportions required a statistical approach suited to compositional data. Because the three behavior categories must sum to the whole of wear time, an increase in one necessarily means a decrease in another, and standard regression techniques that treat each category separately can produce misleading results. The team instead applied Dirichlet regression, a method that preserves the compositional structure of the outcome and directly handles the skewness and heteroscedasticity typical of such data. Parental movement compositions were transformed into additive log-ratios, with sedentary behavior as the reference component, producing two intuitive predictors: the log-ratio of moderate-to-vigorous activity to sitting, and the log-ratio of light activity to sitting. Models were adjusted for the child’s age and sex, parental education, and language region, and the researchers tested whether the associations varied by the child’s age or sex.

The cross-sectional results were consistent across both parents. When mothers replaced sedentary time with moderate-to-vigorous activity, the ratio of that activity to sitting in their children rose in parallel, with a coefficient of 0.10 and a p-value below 0.001; the equivalent paternal coefficient was 0.09 with a p-value of 0.002. Replacing sitting with light activity produced even clearer associations: 0.13 for mothers and 0.09 for fathers, both statistically significant. In plain terms, when mothers or fathers traded sitting for movement of either intensity, their children tended to do the same. Notably, these associations did not depend on the child’s age or sex, suggesting that the link between parental and child movement patterns operates broadly across the six-to-sixteen age range and across both sexes.

The longitudinal findings, though based on a smaller sample, carried the analysis into territory few studies have reached. Five years after the baseline measurement, mothers’ replacement of sitting with light activity still predicted a similar shift in their children, with a coefficient of 0.12 and a p-value of 0.021. The paternal equivalent was directionally similar at 0.10 but fell short of conventional statistical significance at p equal to 0.108, likely reflecting the smaller number of father-child pairs. One intriguing exception emerged: a mother’s shift from sitting to moderate-to-vigorous activity was associated with a larger light-activity response in daughters than in sons, hinting at same-sex modeling dynamics that echo the team’s earlier SOPHYA findings of stronger mother-daughter and father-son activity associations independent of socioeconomic and neighborhood factors.

The effect sizes deserve careful interpretation. When the researchers modeled a parent cutting sedentary time by twenty percent and increasing light activity by the same amount, the predicted change in the child’s light-activity-to-sitting ratio was about eighteen times smaller than the mother’s actual observed shift and about twenty-nine times smaller than the father’s. Children, in other words, move part of the way with their parents but do not mirror them completely. Ternary plots, which display the three movement components as axes of a triangle, illustrated this graphically: parental trajectories from the most sedentary to the least sedentary compositions produced child trajectories in the same direction but of visibly smaller magnitude. The study also found that children’s sedentary share of wear time climbed from about sixty percent at baseline to seventy-three percent at follow-up, underscoring how strongly the adolescent years pull toward sitting.

For public health, the implications point toward family-level intervention design. Because mothers at the low end of sedentary time compensated almost entirely with light activity, while fathers at the low end increased both light and moderate-to-vigorous activity, the authors argue that interventions should not fixate exclusively on vigorous exercise. Light-intensity movement carries documented health benefits for adults and children alike, and the strong, consistent associations within both intensity domains suggest that simply getting parents off the couch and into gentle motion may pay dividends across the whole household. Meta-analytic evidence shows that dyadic interventions targeting shared goals can reduce sedentary behavior, yet programs that address compositional movement patterns remain scarce.

The authors are candid about limitations. The cross-sectional analysis cannot rule out reverse causation, in which active children encourage active parents, and the smaller longitudinal sample limited statistical power, particularly for detecting sex-specific effects among fathers. Sleep was not objectively measured, so the analysis captures only waking behavior rather than the full twenty-four-hour profile, and the findings may not generalize beyond Switzerland’s cultural and socioeconomic context. Attrition bias appears unlikely to distort the results, however, because follow-up participation was not associated with baseline movement composition of either parents or children. The overall message stands: parental movement patterns, and sedentary time in particular, may be effective targets for interventions that benefit parents and children simultaneously, with effects that persist over years rather than fading with the measurement week. As the researchers conclude, integrating family-based strategies into physical activity promotion could yield long-lasting benefits, even if each individual shift is modest.

Subject of Research: Cross-sectional and longitudinal associations between parental and youth movement compositions measured by accelerometry in the Swiss SOPHYA cohort

Article Title: Movement patterns of youth and their parents: cross-sectional and longitudinal associations

Article References: Hänggi, J., Lovison, G., Jeong, A., Dössegger, A., Taube, W., Kayser, B., Suggs, S. L., Bringolf-Isler, B., & Probst-Hensch, N. (2026). Movement patterns of youth and their parents: cross-sectional and longitudinal associations. Journal of Activity, Sedentary and Sleep Behaviors, 5(1), Article 2. https://doi.org/10.1186/s44167-025-00092-w

Image Credits: AI Generated

DOI: 10.1186/s44167-025-00092-w

Keywords: physical activity, sedentary behavior, parent-child dyads, accelerometry, compositional data analysis, SOPHYA cohort, longitudinal study, light physical activity, moderate-to-vigorous activity, youth health, behavioral modeling, Switzerland