In daily work, there is a large number of remote document signing requirements. At present, the mainstream method is to mail the signed documents or transmit pictures of the signed documents. Obviously, mailing some confidential documents is not conducive to confidentiality, and the picture transmission of the signed documents is easy to leak and cannot guarantee the signature of the person. Combined with the above scenarios, a secure remote document signing solution is urgently needed in the current market. Based on the design idea of hardware and software collaboration,we introduce the domestic Hummingbird E203 processor with RISC- V architecture as the core controller, use FPGA (field-programmable gate array) to realize the convolutional neural network handwritten character recognition hardware accelerator and digital watermark embedding algorithm, and carry out functional simulation and experimental data collection.