Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Discovering Behavioral Predispositions in Data to Improve Human Activity Recognition

Maximilian Popko, Sebastian Bader, Stefan Lüdtke, Thomas Kirste

TL;DR

This work tackles the challenge of variable dementia-related behaviors in sensor-based HAR by introducing Behavioral Predispositions (BPDs), which capture the tendency to exhibit certain behaviors in specific time segments. The authors extract BPDs in two ways: a time-based approach and a clustering-based approach that groups segment-wise annotation histograms, with each BPD enabling a dedicated classifier. Empirical results show that when the current BPD is known, HAR performance improves substantially (e.g., up to ~0.5 F1 with 20 BPDs and 30-minute segments using SVM) compared with a baseline majority vote. The study highlights the potential of data-driven priors for dementia care HAR, while noting the practical need to infer BPDs from sensor data and generalize across patients for real-world use.

Abstract

The automatic, sensor-based assessment of challenging behavior of persons with dementia is an important task to support the selection of interventions. However, predicting behaviors like apathy and agitation is challenging due to the large inter- and intra-patient variability. Goal of this paper is to improve the recognition performance by making use of the observation that patients tend to show specific behaviors at certain times of the day or week. We propose to identify such segments of similar behavior via clustering the distributions of annotations of the time segments. All time segments within a cluster then consist of similar behaviors and thus indicate a behavioral predisposition (BPD). We utilize BPDs by training a classifier for each BPD. Empirically, we demonstrate that when the BPD per time segment is known, activity recognition performance can be substantially improved.

Discovering Behavioral Predispositions in Data to Improve Human Activity Recognition

TL;DR

This work tackles the challenge of variable dementia-related behaviors in sensor-based HAR by introducing Behavioral Predispositions (BPDs), which capture the tendency to exhibit certain behaviors in specific time segments. The authors extract BPDs in two ways: a time-based approach and a clustering-based approach that groups segment-wise annotation histograms, with each BPD enabling a dedicated classifier. Empirical results show that when the current BPD is known, HAR performance improves substantially (e.g., up to ~0.5 F1 with 20 BPDs and 30-minute segments using SVM) compared with a baseline majority vote. The study highlights the potential of data-driven priors for dementia care HAR, while noting the practical need to infer BPDs from sensor data and generalize across patients for real-world use.

Abstract

The automatic, sensor-based assessment of challenging behavior of persons with dementia is an important task to support the selection of interventions. However, predicting behaviors like apathy and agitation is challenging due to the large inter- and intra-patient variability. Goal of this paper is to improve the recognition performance by making use of the observation that patients tend to show specific behaviors at certain times of the day or week. We propose to identify such segments of similar behavior via clustering the distributions of annotations of the time segments. All time segments within a cluster then consist of similar behaviors and thus indicate a behavioral predisposition (BPD). We utilize BPDs by training a classifier for each BPD. Empirically, we demonstrate that when the BPD per time segment is known, activity recognition performance can be substantially improved.
Paper Structure (15 sections, 1 equation, 5 figures, 1 table)

This paper contains 15 sections, 1 equation, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Data and Observations
  3. insideDem Project
  4. Observations
  5. Methods
  6. Extracting behavioural predispositions
  7. Time-based BPD
  8. Clustering-based BPD
  9. Classification of Sensor Data
  10. Experimental Evaluation
  11. Experimental Design
  12. Results and Discussion
  13. Future work
  14. Related work
  15. Conclusion

Figures (5)

  • Figure 1: Overview of the concept. We calculate the distribution of annotations for a given segment of the day (histograms). We then cluster these distributions. Next, we use the cluster label (BPD) to select the classifier (color indicates the corresponding classifier). The classifier predicts the behavior from the features of the motion sensor data.
  • Figure 2: Distribution of annotations per person over all days of recordings.
  • Figure 3: Percentual distribution of annotations for one subject and one-hour intervals of the day. (a) shows the average of the annotation and (b) describes the distribution (variance).
  • Figure 4: Averaged $F_1$ score for varying number of clusters and segment lengths. For time-based clustering, the segment length is implicitly given through the number of clusters. For k-means clustering, the segment lengths of the histograms are represented by the color. The segment length is in minutes. Rectangles denote similar BPD structures because of the same number of clusters and similar segment lengths.
  • Figure 5: Confusion matrix of the predicted behaviors using 20 BPDs and thus 20 SVMs. A segment length of 30 minutes were used to calculate the histrogams.