近幾年來，多媒體資料的搜尋與索引的技術受到愈來愈多的注意，尤其是在內涵式搜尋這塊領域中。在音訊這個範圍裡，許多的研究皆專注在MIDI或WAV這些未壓縮的音樂格式上，很少會針對壓縮過的音樂格式這個領域來探討。然而，MP3在目前已是一個重要且受歡迎的音樂格式，許多網路或是個人電腦中所流通的都是以MP3為主。在本篇論文中，我們提出了一個利用以歌找歌方式並且以MP3為主的內涵式音樂搜尋系統。我們將MP3中的次頻帶值取出來當成特徵值，並且利用量化樹索引方法及旋律線擷取方法將所取出的特徵值建成我們要的索引資料，最後再利用這些索引資料作為輸入歌曲片段與資料庫歌曲之相似度比對的依據。 此外，我們將整個內涵式音樂搜尋系統移植到以ARM為主的嵌入式系統平台上來做實現，整個系統的評估實驗分別進行在PC及嵌入式平台上，實驗結果顯示，在整個搜尋的準確性方面，在第一首找到的機率大約在70%左右，大約有90%的機率可以在前三首歌中找到。 In recent years, the searching and indexing techniques for multimedia data are getting more attention in the area of multimedia databases. As many research works were done on the content-based retrieval of MIDI or waveform format music, less attention was received on the compressed domain audio data. As the explosive growth of internet, MP3 audio has become one of the most important and popular media. Most of the audio accessed from database of Internet are all compressed on MP3 domain, not raw data or even low quality MIDI audio. In this thesis, the content-based retrieval of audio example on MP3-based (MPEG 1 layer III) digital music archive is considered. In the proposed approach, the sub-band coefficients (SBC) in a MP3 frame are used. These values are extracted from the MP3 decoder to compute the MP3 features for indexing the MP3 objects. A quantization-tree (QT) and the melody-line tracking (MLT) method are also proposed for indexing MP3 objects. These indexes are used to measure the similarity between MP3 objects. Evaluations on a content-based MP3 retrieval system are performed on the pc version and platform-based embedded system version which indicate the proposed approach can achieve a good performance.
