實驗室經由歷屆學長姐的努力，發展出了一套音樂搜尋系統稱之為CBMR(Content-based Music Retrieval)，在這個系統中，使用者可以透過哼唱甚至不需要歌詞來做查詢的動作。其比對核心，主要是根據DTW(Dynamic Time Warping)來進行比對，這樣的比對方式可以允許使用者唱快或唱慢，因此辨識率上有九成以上的優異表現。

但相對的，DTW卻也有需要大量運算量的缺點，為了能讓這樣的系統應用更廣，資料庫能夠更龐大，全曲比對速度跟辨識率都提升，只有以往的分散式處理是不夠的，最根本的，還是得回歸到單機版的加速，進而再結合平行處理，才能發展出一個夠強大的系統。

因此本論文的重點在找尋一種演算法，此演算法能夠藉著簡單的運算來求得一個近似DTW值的Lower Bound來當做門檻值，此方法主要目的在減少DTW的運算次數而加快DTW的運算速度。有很多相關的加速研究，但是那些演算法都不適用在音樂比對的特性上，因此，我們針對音樂特性提出以下幾點來加以改良：

一. 改變time warping的範圍限制
二. 切段
三. 相關係數
四. 變異數

Many people in my laboratory have been work on a system named CBMR for many years, and we also have pretty good query achievement. However, our system relies on DTW, an algorithm with massive time and space complexity. The recognition rate is also much lower for searching anywhere.

To build an online music search engine, expending the numbers of the MIDIs in the database and increasing the speed and the stability of the searching engine are very important. Because the search time is proportional to the numbers of songs in the database, speed-up becomes a very important presupposition.

This thesis is about finding the speed-up mechanisms suitable for query by singing.

[1] E. Keogh, "Exact Indexing of Dynamic Time Warping," in Proc. of VLDB
Conference, Hong Kong, China, Aug. 20-23, 2002.
[2] Kim, S,. Park, S., & Chu, W. (2001). An Index-based approach for similarity search supporting time warping in large sequence databases. In Proc 17th International Conference on Data Engineering, pp 607-614.
[3] Park, S., Lee, D., & Chu, W. (1999). Fast retrieval of similar subsequences in long sequence databases. In 3rd IEEE Knowledge and Data Engineering Exchange Workshop.
[4] Park, S., Kim, S, & Chu, W. (2001). Segment-based approach for subsequence searches in sequence databases. In Proceedings of the 16th ACM Symposium on Applied Computing, pp. 248-252, Las Vegas, NV, USA.
[5] Park, S., Chu, W., Yoon, J & Hsu, C. (2000). Efficient searches for similar subsequences of different lengths in sequence databases. In Proc. 16th IEEE Int'l Conf. on Data Engineering, pp. 23-32.
[6] Yi, B, K. Jagadish, H & Faloutsos (1998). Efficient retrieval of similar time sequences under time warping. In ICDE 98, pp 23-27.

[7]Simon Sheu , “Effective Filtering for Nearest-Neighbors Queries in Large
Time-Series Databases” , Taiwan , 2003
[8].Teddy Siu Fung Fong, Man Hon Wong(2003), “Efficient Subsequence Matching for Sequences Databases under Time Warping”, inSeventh International Database Engineering and Applications Symposium (IDEAS'03) pp.139
[9] Sanghyun Park, Wesley W. Chu, Jeehee Yoon, Chihcheng Hsu , (2000)“Efficient Searches for Similar Subsequences of Different Lengths in Sequence Databases” 16th International Conference on Data Engineering pp.23
[10] Antonin Guttman. R-tree: a dynamic index structure for spatial searching.
Proc. ACM SIGMOD Conf. on Management of Data, 322-331, 1990.
[11] Y. Zhu, and
D. Shasha. Warping Indexes with Envelope Transforms for Query by Humming. Proc. 2003 ACM SIGMOD
Conf. on Management of Data. pp. 181-192