Boosting fine-grained activity sensing by embracing wireless multipath effects

With a big success in data communication, wireless signals are now exploited for fine-grained contactless activity sensing including human respiration monitoring, finger gesture recognition, subtle chin movement tracking when speaking, etc. Different from coarsegrained body and limb movements, these fine-grained movements are in the scale of millimetres and are thus difficult to be sensed. While good sensing performance can be achieved at one location, the performance degrades dramatically at a very nearby location. In this paper, by revealing the effect of static multipaths in sensing, we propose a novel method to add man-made "virtual" multipath to significantly improve the sensing performance. With carefully designed "virtual" multipath, we are able to boost the sensing performance at each location purely in software without any extra hardware. We demonstrate the effectiveness of the proposed method on three fine-grained sensing applications with just one Wi-Fi transceiver-pair, each equipped with a single antenna. For respiration monitoring, we can remove the "blind spots" and achieve full coverage respiration sensing. For finger gesture recognition, our system can significantly increase the recognition accuracy from 33% to 81%. For chin movement tracking, we are able to count the number of spoken syllables in a sentence at an accuracy of 92.8%.