Title: Lyric alignment in popular songs
1Lyric alignment in popular songs
2Outline
- Project description techniques
- Some important knowledge
- Base-line system
3Outline
- Project description techniques
- Some important knowledge
- Base-line system
4Project description
- Given
- textual transcription of lyrics
- acoustic musical signal
- of a song
- Purpose find the timestamps for the beginning
and ending points for each line of the song.
5Techniques to be investigated
- A repetition based technique for detecting
self-similarities in both the audio and the text - Dynamic programming (also known as dynamic time
warping) will be employed to align the repetition
analysis of both media together to produce an
alignment
6Outline
- Project description techniques
- Some important knowledge
- Base-line system
7Chroma vectors
0.3
0.5
0.1
0.7
8Dynamic time warping algorithm
- Actually, dynamic programming
9Word similarity using phoneme dictionary
- Using CMU phoneme dictionary
- Each word is decomposed into phoneme sequence
- ALIGNMENT AH0 L AY1 N M AH0 N T
- ALGORITHMS AE1 L G ER0 IH2 DH AH0 M Z
- Similarity of 2 words is the similarity of the 2
phoneme sequences
10Outline
- Project description techniques
- Some important knowledge
- Base-line system
11Base-line system overview
Musical signal input
Text input
Music notation sequence
Word sequence
Word self-similarity matrix
Symbol Self-similarity matrix
12Base-line system overview
Musical signal input
Text input
Music notation sequence
Word sequence
Word self-similarity matrix
Symbol Self-similarity matrix
13Base-line system overview
Musical signal input
Text input
Music notation sequence
Word sequence
Word self-similarity matrix
Symbol Self-similarity matrix
14Base-line system overview
Musical signal input
Text input
Music notation sequence
Word sequence
Symbol Self-similarity matrix
Word self-similarity matrix
15Base-line system overview
Musical signal input
Text input
Music notation sequence
Word sequence
Symbol Self-similarity matrix
Word self-similarity matrix
16