本論文的主題是希望透過立體聲音訊之間的關係做為特徵值,建構一個歌曲檢索系統。我們擷取出一些左聲道、右聲道中的峰值(Peak)訊號資訊,以此訊號資訊加以計算成ICTD並做為特徵值,並且利用RLCS以及DTW做為搜尋比對的基礎,來建構出歌曲的檢索系統,最後我們將ICTD特徵值和MPEG-7音訊特徵值(Audio Descriptor)做多重階層的搜尋,來探討這整個歌曲的檢索系統所達成的效果。
In this thesis, we want to use stereo information as characteristic to establish a musical retrieval system. We extract the peak information in left channel and right channel, using this peak information to calculate ICTD as characteristic, and then compare these characteristic by RLCS and DTW algorithm to establish musical retrieval system. Finally, stereo information will do multi level search with MPEG-7 Audio Descriptor to explore the entire music retrieval system.