 This research demonstrates the potential of AMODE ultrasound technology for gesture recognition. It uses a human-machine interface, HMI, to interact with users in real-time, and trains machine learning models using data processing techniques such as Gaussian filtering, feature extraction, and principle component analysis, PCA. The NB, LDA, and SVM algorithms are used to classify gestures which are then implemented in C++ to achieve real-time recognition. The results of both offline and real-time experiments indicate that this system achieves high accuracy and efficiency. This article was authored by Zongxing Liu, Xiaxiong Kai, Bingxing Chen, and others.