Towards realtime measurement of connectedness in human movement

With the proliferation of wearable sensors, we have access to rich information regarding human movement that gives us insights into our daily activities like never before. In a sensor rich environment, it is desirable to build systems that are aware of human interactions by studying contextual information. In this paper, we attempt to quantify one such contextual cue - the connectedness of physical movement. Inspired by the Semblance of Typology Entrainments, we estimate the connectedness of trained dancers as observed from inertial sensors, using a diverse set of techniques such as quaternion correlation, approximate entropy, Fourier temporal pyramids, and discrete cosine transform. Preliminary experiments show that it is possible to robustly estimate connectedness that is invariant to frequency, amplitude, noise or time lag.