How many kinects should look at you? A multi-agent system approach

Abstract

This work compares the combined use of several infrared sensors such as the Microsoft Kinect. The system is modeled as a multi-agent system where each Kinect sensor is considered a physical/intelligent agent provided with pattern recognition capacity. The proposal offers a set of experiments that study when under concrete and different scenarios the combination of agent sensors achieves more accurate results than a single agent. So, the experiments analyze the accuracy and data resolution obtained from Microsoft Kinect sensors. Concretely, six experiments are conducted and the results are analyzed with different configurations of a variable number ofMicrosoft Kinect sensors. The results indicate that a single Kinect agent provides a higher accuracy in comparison to more Kinects. However, additional agents can be considered to increase the flexibility of the capture system.