Machine Learning Based Signal Detection for Ambient Backscatter Communications

Abstract

The ambient backscatter communication (AmBC) system enables radio-frequency (RF) powered devices (e.g., tags, sensors) to transmit their information bits to readers by backscattering and modulating the ambient RF signal. Different from traditional radio-frequency identification (RFID) systems, an AmBC system does not require a reader to transmit excitation signals to the tag and there is no additional carrier emitters required. Therefore, AmBC systems exhibit low-cost and high energy efficiency. The existing AmBC systems utilize an energy detector or a Minimum Mean Square Error (MMSE) detector to detect tag signals which suffers from high bit error rate (BER). In this paper, a machine learning based detection method is proposed to detect the tag signals for an AmBC system by transforming the detection problem into a classification problem. In more detail, the proposed method classifies the received signals into two groups based on the energy features of the received signals. Our simulation results show that the proposed machine learning based detection method outperforms the traditional detection methods, especially in the low SNR regime.