Compressive infrared sensing for arm gesture acquisition and recognition

Abstract

This article reports the findings of a compressive infrared sensing approach for arm gesture acquisition and recognition. The spatial-temporal changing motion information are intrinsical clues for determining the semantics of gesture, which can be considered as a sparse spatial distribution compared with the sensing region. We first built a database of dynamic arm gestures with writing Arabic numerals 0 – 9 in a constrained interaction region, and study its sparse property of spatial distribution. Then we design a pyroelectric infrared (PIR) sensor array with random visibility modulation for compressive gesture acquisition. The semantic recognition of gesture is executed directly on the sensors' low-dimensional sequence using vector quantization (VQ) technology. The experimental results demonstrate the effectiveness of the proposed sensing method.