支持得分矩阵的音乐检索技术

Fast and sensitive matching of music using position specific scoring matrices

作为音乐检索的重要方式,哼唱检索由于其有效性和方便性,引起了广泛的关注。对此提出了一种新的基于得分矩阵的音乐哼唱快速检索技术,可以实现哼唱音乐的快速检索。首先根据哼唱音乐特征,将音乐数据库和用户提供的哼唱片段,按自然停顿方式划分音乐的语句,同时使用K-means聚类算法对音乐的语句片段进行音高相似性计算,并根据聚类情况提取出位置特异性得分矩阵。此外,基于得分矩阵提出NA匹配算法和两种加速分段计分方法,分别是顺序前瞻计分SLS算法和置换矩阵前瞻计分PLA算法。实验结果表明所提出的基于得分矩阵的音乐检索技术能够快速有效地返回查询结果,同时PLA算法具有更有效的哼唱音乐检索结果。

As an important way of music retrieval, humming query has gained wide attention because of its effectiveness and convenience. This paper proposes a novel retrieval technique based on the scoring matrices of humming, which can provide fast retrieval for humming query. In the proposed technique, the music database and humming given by users are first partitioned according to natural pauses, and k-means clustering algorithm is adopted to compute pitch similarity. This paper sets specific scoring matrix according to clustering. Based on scoring matrix, this paper further proposes a brute force pattern matching algorithm, as well as two accelerated methods. The experimental results demonstrate both the efficiency and effectiveness of the retrieval method proposed in this paper, and the PLA algorithm has more effective humming music retrieval result.