基于空间自适应哈希算法的并行双音频指纹技术  

Parallel dual audio fingerprinting based on spatially adaptive hashing algorithms

作　　者：黄一英 邓开发[2] 邬春学[1] Huang Yiying;Deng Kaifa;Wu Chunxue(School of Optical-Electrical&Computer Engineering,University of Shanghai for Science&Technology,Shanghai 200093,China;School of Art&Design,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093 [2]上海工程技术大学艺术设计学院,上海201620

出　　处：《计算机应用研究》2022年第7期2076-2080,2086,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(2018YFC0810204)。

摘　　要：针对非法音频拷贝的检索跟踪性能问题进行研究,提出了一种基于新哈希的并行双音频指纹方法。为了提高检测片段的完整性,应用了并行的特性,通过沃尔什—哈达玛转换生成两组实时并行的指纹;通过新哈希方法对两组指纹进行相似性计算;最后检索时让并行的两组指纹进行匹配,合并得到最终结果。实验结果表明,该方法不仅对采用不同方法处理的音频具有良好的鲁棒性,能稳定地表征音频整体特性,与所考虑的其他方法相比,所提出的方法在匹配段完整性方面有所提高。Aiming at the retrieval and tracking performance of illegal audio copies,this paper proposed a new hash-based parallel dual audio fingerprinting method.Firstly,to improve the integrity of the detected fragments,it applied the parallel feature and Walsh-Hadamard transformation to generate two sets of real-time parallel fingerprints.Then,it computed the similarity between the two sets of fingerprints by the new hashing method.Finally,during retrieval,it matched the parallel two sets of fingerprints and merged to obtain the final result.The experimental results show that the proposed method has good robustness for audio adopted different methods,and can stable characterize the overall characteristics of audio.Compared with the other methods,the proposed method has improved in terms of matching period of integrity.

关 键 词：音频指纹 并行双音频 沃尔什—哈达玛转换 空间自适应哈希算法 频谱分割 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]