Window Size Impact in Human Activity Recognition

Abstract

Signal segmentation is a crucial stage in the activity recognition process; however, this has been rarely and vaguely characterized so far. Windowing approaches are normally used for segmentation, but no clear consensus exists on which window size should be preferably employed. In fact, most designs normally rely on figures used in previous works, but with no strict studies that support them. Intuitively, decreasing the window size allows for a faster activity detection, as well as reduced resources and energy needs. On the contrary, large data windows are normally considered for the recognition of complex activities. In this work, we present an extensive study to fairly characterize the windowing procedure, to determine its impact within the activity recognition process and to help clarify some of the habitual assumptions made during the recognition system design. To that end, some of the most widely used activity recognition procedures are evaluated for a wide range of window sizes and activities. From the evaluation, the interval 1–2 s proves to provide the best trade-off between recognition speed and accuracy. The study, specifically intended for on-body activity recognition systems, further provides designers with a set of guidelines devised to facilitate the system definition and configuration according to the particular application requirements and target activities.

Enabling process mining on sensor data from smart products

“…The optimal size of the individual windows depends mainly on the sampling rate of the sensor and the expected duration of activities [12]. There is a trade-off between having enough measurements to accurately detect the type of behaviour in a window and the possibility of having multiple changes in behaviour within a window.…”

“…These approaches vary in complexity and in their assumptions on the input data. Most activity recognition approaches take as their input a set of time windows, segments of sensor data, or discrete events with certain features that are then classified to identify which activities were performed during these observations [12]. However, this means that these techniques often need large amounts of training data in order to learn a classifier to recognise the activities [13].…”

“…Various approaches from the field of activity recognition produce such mappings between segments of sensor data and recognisable activities [12]- [15]. These approaches vary in complexity and in their assumptions on the input data.…”

Combining Human Action Sensing of Wheelchair Users and Machine Learning for Autonomous Accessibility Data Collection

“…Overlap ratio is the fraction of overlapping between S k−1 and S k , assuming that S k is the k-th segment. Although window size affects classification accuracy [32], the selection of a size is difficult and requires experimental evaluation. To find an optimal window, we tested different window sizes from 50 to 1,000 and investigated the relationships among the effects of window size.…”

Recognition of Daily Human Activity Using an Artificial Neural Network and Smartwatch

“…The sliding window technique is widely used and has been proven effective for handling streaming data [16,17]. share parts of the sensor readings.…”