Cushionware: A practical sitting posture-based interaction system

Sitting posture-based interaction is to leverage user's sitting posture as an input for interaction. In order to realize it, user's sitting posture needs to be recognized accurately. However, existing works on sitting posture recognition either use intrusive wearable/visual sensors or rely on expensive high-resolution pressure sensor array, and thus hindering the widespread adoption. In this work, we introduce Cushionware, a practical sitting posture recognition system that is based on sparse pressure sensor array. Pressure sensor array is placed within the chair cushion to collect pressure data while user is sitting. After collecting the data, we first model sitting posture by extracting a set of user-invariant features and then identify the sitting posture using machine learning method. To demonstrate the utility of Cushionware, we develop two applications including video game playing and wheelchair motion control.