Modulation mode identification method based on FPGA and deep learning

The invention discloses a modulation mode identification method based on an FPGA (Field Programmable Gate Array) and deep learning, relates to the modulation mode identification method based on the FPGA and the deep learning, and aims to realize modulation mode identification on an FPGA platform by using a lightweight Transformer and realize high precision, low delay and low power consumption under a relatively complex modulation mode. The method comprises the following steps: obtaining a trained deep learning model, and exporting a weight and an offset; a Vivado HLS tool is used for building an algorithm of a deep learning model at a PL end of the PYNQ-Z2, and a compiled code is obtained; generating an IP core of the HDL through the compiled code; importing the IP core into Vivado to obtain a. Bit file, a. Tcl file and a. Hwh file, and storing the. Bit file, the. Tcl file and the. Hwh file; storing the data to be tested, the weight, the bias and the stored file into an SD (Secure Digital) card of PYNQ-Z2; st