In this study, a series of experiments were conducted to examine the feasibility of using Wi-Fi signals for gesture recognition in indoor environments. Totally different from traditional sensor-based approach and computer vision approach, the proposed novel gesture recognition method does not require line-of-sight and sensors placed in the body. Furthermore, unlike recent work using Doppler shifts based on Wi-Fi Radar, the proposed method can be achieved only by Wi-Fi transmissions between a transmitter and a receiver, which can transmit the information and recognize gesture simultaneously. We evaluate the proposed method using Sora platform in an office environment, with a human subject performing eight different gestures. The type of the gestures performed between the transmitter and receiver of Sora platform can have significant effects on the received Wi-Fi signals. From these time varying signals, we extract features that are representative of the gesture types based on 1-D diagonal slice of fourth-order cumulants within a sliding time window. Then, we use support vector machine (SVM) to realize the gesture recognition. Our results show that proposed method can recognize a set of eight gestures with an average accuracy of 96.44%.