基于频域和时域差分的音频指纹算法研究  

Research on Audio Fingerprinting Algorithm Based on Difference in Frequency and Time Domain

作　　者：牛宪华[1] 曾柏森[2] 陈思利[3] 

机构地区：[1]西华大学数学与计算机学院,四川成都610039 [2]中国联通成都市分公司网络优化中心,四川成都610036 [3]成都工业学院通信工程系,四川成都611730

出　　处：《西华大学学报（自然科学版）》2014年第5期10-15,共6页Journal of Xihua University:Natural Science Edition

基　　金：四川省教育厅自然科学项目(13ZB0027);四川省高校重点实验室开放研究基金项目(szjj2013-023)

摘　　要：针对现有算法在某些信号畸变情况下鲁棒性不理想的问题,修改了频带滤波器组模块的功率谱子带划分方式和二维滤波器模块的子带间时频差分方式,分别提出了基于频域和时域差分的2种音频指纹改进算法。实验结果表明,改进算法在低信噪比、线性速度改变等方面的鲁棒性均优于现有的算法。This paper is concerned with the analysis of audio fingerprint algorithm based on Fourier transform .In allusion to occur-ring problems of the existing algorithm in robustness , the authors modified the mode of energy division and difference between the sub-bands and presented an improved audio fingerprint algorithm which applied the difference in frequency and time domain .The pro-posed algorithm was analyzed subsequently , and experimental results showed that the proposed algorithms improved the robustness in several aspects such as low SNR and linear speed change than existing algorithm .

关 键 词：音频指纹 功率谱 鲁棒性 

分 类 号：TN912.34[电子电信—通信与信息系统] TP391.42[电子电信—信息与通信工程]