sensor-modeling

sensor-modeling is a research toolkit for behavioural sensor streams in ambient assisted living, digital health and smart-home studies. It runs an end-to-end pipeline from heterogeneous sensor observations to explained alerts, on top of a modelling core of Bernoulli autoregressive models, hidden Markov models, change-point detection and non-homogeneous Poisson processes. The methods are chosen for the awkward properties of ambient data: irregular sampling, frequent missingness, binary activations, and the need for models a clinician can read.

It is a research toolkit, not a medical device. Nothing it produces is a diagnosis, and the quantitative results in its documentation come from the bundled simulator rather than validated real-world data.

Install

1
pip install sensor-modeling

Package Metadata

  • Current release: 0.2.0
  • Requires Python: >=3.10,<3.13
  • Status: beta

Where It Fits

Use it for reproducible analysis of activity and presence sensors where interpretability matters more than raw accuracy: routine modelling, anomaly and change-point detection in daily patterns, and privacy-preserving multimodal fusion. It is the software companion of the behavioural-sensing research listed on the papers page.

Loading mathematical content