A 24 GHz frequency modulated continuous wave radar system to recognize human's hand gestures is implemented, which uses commercial off-the-shelf RF front-end IC with one transmitter and four receivers. Planar patch array antennas, signal conditioning circuits and interconnections to a PC are designed for the system. Range-Doppler maps for four receiver channels are obtained with saw-tooth chirping signals transmitted to detect hand gestures. The radar system shows real-time highly accurate gesture recognition. Long-short term memory recurrent neural network as a supervised machine learning technique is used. Seven kinds of hand gestures are recognized within 0.4 m and ±30° from the center of the transmitted antenna with above 91 % accuracy.