Unsupervised adaptation for acceleration-based activity recognition: robustness to sensor displacement and rotation

Abstract

A common assumption in activity recognition is that the system remains unchanged between its design and its posterior operation. However, many factors affect the data distribution between two different experimental sessions. One of these factors is the potential change in the sensor location (e.g. due to replacement or slippage) affecting the classification performance. Assuming that changes in the sensor placement mainly result in shifts in the feature distributions, we propose an unsupervised adaptive classifier that calibrates itself using an online version of expectationmaximisation. Tests using three activity recognition scenarios show that the proposed adaptive algorithm is robust against shift in the feature space due to sensor displacement and rotation. Moreover, since the method estimates the change in the feature distribution it can also be used to roughly evaluate the reliability of the system during online operation.

The Opportunity challenge: A benchmark database for on-body sensor-based activity recognition

“…Null class) (Stiefmeier et al, 2008). For these reason robust methods are required tackling issues ranging from the feature selection and classification (Preece et al, 2009), to decision fusion and fault-tolerance (e.g., Chavarriaga et al (2012); Sagha et al (2011b); Zappi et al (2007)). …”

Kinect=IMU? Learning MIMO Signal Mappings to Automatically Translate Activity Recognition Systems across Sensor Modalities

“…Self-calibration approaches require no user intervention, but were demonstrated only for specific cases (e.g. displacement of accelerometers [3], [4]). Combinations of multiple sensor modalities also allow to tolerate on-body displacement [5], or to substitute sensor modalities [6].…”

Adaptive Gesture Recognition with Variation Estimation for Interactive Systems

“…They typically require cumbersome training procedures. Consequently, several authors propose so-called adaptive systems, where the system adapts to variability of input, user [Licsár and Szirányi 2005;Wilson 2000;Caridakis et al 2009], or sensor location [Chavarriaga et al 2013]. These systems are fundamentally not designed to take into account variations that occur during the movement performance.…”