Raising the Bar(ometer): Identifying a User's Stair and Lift Usage Through Wearable Sensor Data Analysis

TL;DR

The paper tackles differentiating stair climbing from elevator use using wearable sensor data and introduces a new dataset collected from 20 participants with accelerometer and barometer inputs. A Random Forest with 8-second windows achieves a high accuracy (87.61%) and strong multiclass F1 performance (approximately 0.87) across five classes, including Null, Stairs Up/Down, and Lift Up/Down. An ablation study shows pressure data critically enhances discrimination, with macro F1 dropping markedly when pressure is removed. The work highlights the value of barometric sensors in activity recognition, identifies slope_pressure as a key feature, and demonstrates the impact of window size on performance, contributing a publicly available lift-vs-stair dataset and informing health, fitness, and safety applications in pervasive computing contexts.

Abstract

Many users are confronted multiple times daily with the choice of whether to take the stairs or the elevator. Whereas taking the stairs could be beneficial for cardiovascular health and wellness, taking the elevator might be more convenient but it also consumes energy. By precisely tracking and boosting users' stairs and elevator usage through their wearable, users might gain health insights and motivation, encouraging a healthy lifestyle and lowering the risk of sedentary-related health problems. This research describes a new exploratory dataset, to examine the patterns and behaviors related to using stairs and lifts. We collected data from 20 participants while climbing and descending stairs and taking a lift in a variety of scenarios. The aim is to provide insights and demonstrate the practicality of using wearable sensor data for such a scenario. Our collected dataset was used to train and test a Random Forest machine learning model, and the results show that our method is highly accurate at classifying stair and lift operations with an accuracy of 87.61% and a multi-class weighted F1-score of 87.56% over 8-second time windows. Furthermore, we investigate the effect of various types of sensors and data attributes on the model's performance. Our findings show that combining inertial and pressure sensors yields a viable solution for real-time activity detection.

Figures (7)

• Figure 1: The distribution of the activity classes contained in the recorded dataset. The Null class makes up 52 %, while the rest of the data consists of the four target classes (Lift Up, Lift Down, Stairs Up, Stairs Down).
• Figure 2: Example accelerometer and barometer data for the 'Stairs Up' activity class. The accelerometer shows the typical patterns of walking. Meanwhile, the pressure steadily decreases, with short plateaus occurring on the stairway's landings. Additionally, the pressure sensor shows some spikes which are likely caused by the movement of the arm while walking.
• Figure 3: Example accelerometer and barometer data for the 'Lift Up' activity class. While the lift is moving, the participant mostly stood still. We observe the pressure decreasing smoothly.
• Figure 4: Example accelerometer and barometer data for the 'Stairs Down' activity class. The accelerometer shows a lot of activity, with similar walking patterns as in 'Stairs Up'. In direct comparison, the magnitude and frequency of the movement sensors seem higher. The barometer detects the steadily rising pressure, also interrupted by short plateaus caused by the landings.
• Figure 5: Example accelerometer and barometer data for the 'Lift Down' activity class. The specific participant for this instance was standing very still during the descent. We can see the pressure increasing linearly and smoothly.