Indoor non-rhythmic human motion classification using a frequency-modulated continuous-wave radar

Abstract

Human motion classification is widely used in intelligent house, surveillance, search and rescue operation, intelligent house, and elder monitoring. In this study, a frequency-modulated continuous-wave radar is utilised to classify non-rhythmic human motion in an indoor scenario. Both the range and Doppler features are extracted from echo signals for a machine learning classifier subspace K-nearest neighbour. Extensive experiments demonstrate its feasibility, and an accuracy rate of 94.2% was achieved in recognition of eight typical non-rhythmic motions.

Keywords

• cw radar
• feature extraction
• learning (artificial intelligence)
• fm radar
• pattern classification
• doppler radar
• indoor environment
• echo
• indoor nonrhythmic human motion classification
• frequency-modulated continuous-wave radar
• indoor scenario
• range feature extraction
• doppler feature extraction
• echo signals
• machine learning classifier subspace k-nearest neighbour
• nonrhythmic motion recognition