Temporal Pyramid Pooling Convolutional Neural Network for Cover Song Identification

Abstract

Cover song identification is an important problem in the field of Music Information Retrieval. Most existing methods rely on hand-crafted features and sequence alignment methods, and further breakthrough is hard to achieve. In this paper, Convolutional Neural Networks (CNNs) are used for representation learning toward this task. We show that they could be naturally adapted to deal with key transposition in cover songs. Additionally, Temporal Pyramid Pooling is utilized to extract information on different scales and transform songs with different lengths into fixed-dimensional representations. Furthermore, a training scheme is designed to enhance the robustness of our model. Extensive experiments demonstrate that combined with these techniques, our approach is robust against musical variations existing in cover songs and outperforms state-of-the-art methods on several datasets with low time complexity.

Accurate and Scalable Version Identification Using Musically-Motivated Embeddings

“…Results on Da-TACOS 2DFTM [17] 0.275 155 SiMPle [18] 0.332 142 Dmax [14] 0.322 132 Qmax [10] 0.365 113 Qmax* [30] 0.373 104 EarlyFusion [12] 0.426 116 LateFusion [14] 0.454 177 MOVE w/ d = 4 k (ours) 0.489 43 MOVE w/ d = 16 k (ours) 0.506 42 Results on YTC SiMPle [18] 0.591 8 2DFTM sequences [29] 0.648 8 InNet [19] 0.660 6 SuCo-DTW [31] 0.800 3 CQT-TPPNet [20] 0.859 3 MOVE w/ d = 16 k (ours) 0.885 3 Table 2. Comparison of state-of-the-art VI systems (best results are highlighted in bold).…”

Learning a Representation for Cover Song Identification Using Convolutional Neural Network

“…Each song in Youtube has 7 versions, with 2 original versions and 5 different versions and thus results in 350 recordings in total. In our experiment, we use the 100 original versions as references and the others as queries following the same as [15,9,8].…”

“…Results on Youtube DPLA [2] 0.525 0.132 9.43 2420s SiMPle [15] 0.591 0.140 7.91 18.7s Fingerprinting [16] 0.648 0.145 8.27 -SuCo-DTW [17] 0.800 0.180 3.42 4.59s Ki-CNN [8] 0.656 0.155 6.26 0.35ms TPPNet [9] 0.859 0.188 2.85 0.04ms CQT-Net 0.917 0.192 2.50 0.04ms Results on Covers80 NCP-WIDI [18] 0.645 ---CRP [3] 0.544 0.061 --Fusing [19] 0.625 0.071 --Ki-CNN [8] 0.506 0.068 16.4 0.55ms TPPNet [9] 0.744 0.086 6.88 0.06ms CQT-Net 0.840 0.091 3.85 0.06ms Results on Mazurkas DTW [15] 0.882 0.949 4.05 -NCD [20] 0.767 ---Compression [21] 0.795 ---Fingerprinting [22] 0.819 ---SiMPle [15] 0.880 0.952 2.33 -SuCo-repeat [17] 0.850 0.940 2.77 -2DFM [4] 0 Table 1. Performance on different datasets (-indicates the results are not shown in original works).…”

“…Moreover, the query time shown in the table does not include the time of feature extracting. Therefore, our method has the same time consumed as [9]. It extracts a fixeddimensional feature whatever the duration of input audio is.…”

“…Moreover, deep learning approaches are introduced to cover song identification. For instance, CNNs are utilized to measure the similarity matrix [6] or learn features [7,8,9,10]. While these methods have achieved promising results, there is still room for improvement.…”