Over　the　past　decades,　Human　action　recognition　is　very　significant　for　human　action　analysis,　which　is　a　strongly　active　research　area.　The　relevant　target　detection　and　classification　technologies　have　become　research　hot　topics,　with　a　very　wide　range　of　applications,　including　security　protection,　disaster　relief,　smart　home　and　other　fields.　In　the　academia,　radar　has　gradually　become　the　focus　of　research　in　the　field　of　target　detection　and　classification.　Compared　with　the　other　sensors,　radar　is　more　advanced　in　non-environmental　impact　and　good　data　integrity　in　human　body　detection　and　classification.　While　the　current　studies　have　combined　machine　learning　with　radar　signals,　the　performance　of　machine　learning　methods　such　as　SVM　and　decision　tree　is　not　well.　To　address　this　issue,　this　work　utilized　convolutional　neural　networks,　combined　with　micro-Doppler　image　features　extracted　from　radar　signals.　The　experimental　demonstrated　that　CNN　model　achieved　a　higher　accuracy　in　human　activity　recognition　task.　©　2022,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.