Novel bispectrum-based wireless vision technique using disturbance of electromagnetic field by human gestures
Oleh Viunytskyi, Alexander V. Totsky · 2017
Novel human gesture recognition and classification technique is suggested and experimentally studied. Suggested strategy is based on exploiting the interactions of human gestures with high-frequency electromagnetic field. Extracting of classification features contained in the wireless radio signal modulated by human gestures is proposed by utilizing bispectrum-based processing of the signal envelope. Novel two kinds of human gesture classification features are computed in the form of biphase values contained in the given slice on the bispectral plane or array of biphase samples contained within the limits of the main bispectral triangular area. It is shown that phase bispectrum contains information about the shape of wireless signal envelope and, consequently, about type of human gesture. Performance of the bispectrum-based human gesture classification technique is studied experimentally. Experimental results obtained by developed measuring hardware and designed software are represented and discussed. The results obtained for the set of test human gestures demonstrate that proposed technique is non-sensitive to random wireless signal delays and magnitude variations commonly observed in closed multi-path interference environment.