In this paper, a smart antenna array based on the image-aware technique that can automatically switch its beam to track users is proposed. Mobilenet, a deep neural network (DNN), is utilized to find the users' position information from images; it is trained using the collected dataset, and the parameters are optimized. The modified DNN is then implanted into a miniaturized embedded system RK3399. To obtain beam switchable radiation, a 2×2 microstrip patch antenna array with a 4×4 Butler feeding matrix is designed in the 2.4 GHz ISM band. Then, the RK3399 is compactly integrated with the patch antenna array, and a microcamera is used as the eye of the antenna to capture the environment information. A series of measurements of the received power and error vector magnitude (EVM) are performed in a real indoor scenario to verify the design. The results agree well with the expected results and show that the design is capable of guaranteeing communication quality and improving energy efficiency, making this technique a good candidate for indoor wireless applications. INDEX TERMS Automatic beam switching, butler matrix, deep neural network, error vector magnitude, smart antenna.

No abstract

No abstract

No abstract

No abstract