Application Of LDPC Decoding Algorithm Based On Deep Learning In Speech Copyright Protection Technology

With the popularization and development of the mobile internet,media service models continue to innovate.However,this has also brought about serious issues of piracy and copyright infringement,posing threats to the interests of quality content creators and various stakeholders.To address this issue,measures need to be taken to protect intellectual property rights and the interests of content creators.The academic community has proposed techniques such as digital watermarking,blockchain-based evidence storage,Digital Image Correlation(DCI),and Digital Rights Management(DRM).The purification of cyberspace and the protection of intellectual property rights have been effectively promoted and implemented.The ultimate goal of digital copyright protection is to thoroughly curb piracy,but due to technological limitations,infringement of audio content copyright still exists.Therefore,it is necessary to further strengthen technical means and legal supervision.This paper improves the technology of speech copyright protection through the research of Low Density Parity Check(LDPC)codes and speech copyright protection technology,combining LDPC decoding and chaotic encryption theory.This research innovates based on existing technologies,enhancing the reliability and security of speech copyright protection technology,and further proves through experimental simulation that the theory proposed in this paper is feasible and practical compared to existing algorithms.The main work of this paper includes:(1)Improving the existing Log Likelihood Ratio-Belief Propagation(LLR-BP)decoding algorithm.As the appearance of oscillating error bits in LLR-BP decoding adversely affects the decoding algorithm,this paper proposes a smoothing computation method to reduce the negative impact of oscillation effects on decoding and enhance the decoding performance of medium to long-length LLR-BP decoding algorithms.(2)Utilizing deep learning methods to improve the Belief Propagation(BP)decoding algorithm.Since the BP decoding algorithm for medium and short length codes is prone to produce short cycle limitations,and the decoding performance of the BP decoding algorithm decreases when there is a short cycle in the corresponding Tanner graph,this paper studies the process of mapping from the BP algorithm model to the deep learning BP model.With the classification ability of the deep learning network,the correct decoding of LDPC codes is achieved,overcoming short cycle limitations.This is verified through simulation.For high-rate and medium to short-length LDPC codes,the decoding performance of this method significantly improves compared to the traditional BP decoding algorithm.(3)The paper researches speech copyright protection technology,uses deep learning decoding algorithms to decode copyright information encrypted by chaos,and then restores the original copyright information through chaotic decryption.This improves the reliability and security of speech copyright protection technology,and realizes the protection of speech copyright.(4)Firstly,the paper briefly describes the related knowledge of Field Programmable Gate Array(FPGA)hardware and Verilog HDL language.Then,it conducts hardware simulation of the deep learning BP decoding algorithm,analyzes the advantages and disadvantages of the decoder structure,and carries out overall design and modular design.This research enhances the decoding speed and performance of LDPC codes through hardware design.During the implementation process,this research fully utilizes the high programmability and flexibility of FPGA and adopts the Verilog language for design and development.Finally,the feasibility of the algorithm is verified through timing simulation diagrams,proving the algorithm’s practical significance.