摘要： 由于数字音频数据量极大的特点，采用传统音频检索方法会导致等待时间过长。为加快音频检索时间，提出一种基于GPU加速的数字音频检索方法。利用数字音频的特征将连续的音频划分成等长的多个短时音频段，采用GPU加速算法计算每个短时音频段的特征值，将各段的特征值构成特征矩阵。使用后缀数组的变形算法找出2个特征值序列的公共特征段落集合，并将公共特征段落集合进行精化和整体匹配，从而得出检索结果。实验结果表明，该检索方法的准确率可以达到95%以上，与已有方法相比，可以大幅度地提高检索速度，加速比可以达到10倍以上。

关键词: 音频检索, GPU加速, 后缀数组, 音频特征, 特征值序列, 整体匹配

Abstract: As digital audio has a feature of great data volume, traditional audio retrieval method results are used in intolerable response time. In order to speed up audio retrieval, this paper proposes a GPU acceleration audio retrieval method. The audio is divided into multiple short audio segments based on the features, and the characteristic matrix is constituted by the eigenvalues which is calculated from each short audio segment using the GPU acceleration algorithm. The suffix array deformation algorithm is used to find the common set from the two eigenvalues sequence. The common set is refined and overall matched to get the retrieval result. Experimental results show that the retrieval accuracy is over 95% and compared with existing algorithms, this method can significantly improve the retrieval speed and speedup can be achieved in more than 10 times.

Key words: audio retrieval, GPU acceleration, suffix array, audio characteristic, eigenvalue sequence, overall match

中图分类号: 

• TP391