Efficient K-NN search in polyphonic music databases using a lower bounding mechanism

Abstract

Querying polyphonic music from a large data collection is an interesting and challenging topic. Recently, researchers attempt to provide efficient techniques for content-based retrieval in polyphonic music databases where queries can also be polyphonic. However, most of the techniques do not perform the approximate matching well. In this paper, we present a novel method to efficiently retrieve k music works that contain segments most similar to the user query based on the edit distance. A list-based index structure is first constructed using the feature of the polyphony. A set of candidate approximate answers is then generated for the user query. A lower bounding mechanism is proposed to prune these candidates such that the k answers can be obtained efficiently. The efficiency of the proposed method is evaluated by real data set and synthetic data set, reporting significant improvement over existing approaches in the response time yielded.

Exploring composite acoustic features for efficient music similarity query

“…If the child node is an internal node, it means that there is a further partitioning of the space into sub-clusters, and we insert the node into the queue(lines 10-11). Otherwise, the child must be a leaf node, we access the real data points in the node and compute their distances to the query point; points that are nearer to the query point are then used to update the current KNN list (lines [12][13][14][15]. The function Adjust() in line 16 updates the value of pruning distance when necessary, which is always equal to the distance between the query point and the K-th nearest neighbor candidate.…”

“…The first class adopts symbolic representation to keep track of musical information such as tone, pitch and duration. Thus, the problem of music search can be transformed into approximate string matching [15]. Another stream is to convert the music into indexable items, typically points in a high-dimensional vector space [19,21,14].…”

Compacting music signatures for efficient music retrieval

“…In the example, the window sliding step size is 2 beats, and hence the music piece can be represented by four segments, (8,10,8,13,12,12), (8,13,12,12,8,10), (12,12,8,10,8,15) and (8,10,8,15,13,13); in other words, 4 points in a 6-dimensional space. It is possible for us to conduct music retrieval using multi-dimensional index structures.…”

“…Content-based music retrieval has attracted much research interest recently [8,12,13,14,15,18,21,22,23,3,2]. Several content-based retrieval techniques have been proposed in the literature, and they can be classified into two categories, i.e.…”

“…The second class adopts symbolic representation based on the musical scores and notes in the score to keep track of musical information such as tone, pitch and duration. In this case, the problem of music retrieval can be transformed into approximate string matching or time sequence/series data matching [9,10,15,18,25,26,30,31]. Unfortunately, these algorithms are computationally expensive (and hence impractical) for large music databases as the entirety of a database has to be scanned to find matching sequences for each query.…”