论文部分内容阅读
提出一种新颖的基于音符分割和模糊分类的音乐内容认证方法.该算法打破了传统音频认证所采用的固定长度分割方式,将音乐信号分割成一系列具有完整语义信息的不等长音符片段作为认证的基本单元,结合动态时间规整DTW(dynamic time warping)对齐技术,有效解决了大多数现有算法都存在的对同步失真脆弱的问题.在每个音符片段计算基于半阶音符类Chroma的鲁棒Hash值,根据原始音乐与待认证音乐之间Hash值差异的统计特性和时间分布特性,对3种新定义的度量指标进行模糊分类从而得到最终的认证结果.对于未通过认证的音乐信号系统还可以进行篡改区域的检测.实验结果表明,该算法能够有效区分可容许操作和恶意篡改,同时在篡改定位方面具有较高的精度.
This paper proposes a novel music content authentication method based on note segmentation and fuzzy classification.The algorithm breaks the fixed length segmentation method used in traditional audio authentication and divides the music signal into a series of unequal-length musical fragments with complete semantic information as authentication , Combined with the dynamic time warping (DTW) alignment technique, effectively solves the problem of most of the existing algorithms that are vulnerable to synchronization distortions.Calculation based on the Chroma of the half-note class Chroma is calculated for each note segment Hash value, according to the difference between the original music and the music to be authenticated Hash value of the statistical characteristics and time distribution of the three newly defined metrics fuzzy classification to get the final certification results.For the music signal system failed certification Which can detect the tampered area.The experimental results show that the algorithm can effectively distinguish between allowable manipulation and malicious tampering and at the same time has higher accuracy in tamper localization.