Method and apparatus for automatically segueing between audio tracks
There is disclosed a method and storage media for automatically segueing between a first audio track and a second audio track. The first audio track and the second audio track may be analyzed to determine a preferred transition window. Segueing from the first track to the second track may occur during the preferred transition window.
1. A method for music presentation, comprising:
resampling acoustic metadata for a first track at a rhythmic rate of the first track
resampling acoustic metadata for a second track at a rhythmic rate of the second track
performing a sliding window comparison of the resampled acoustic metadata for the first audio track and the resampled acoustic metadata for the second audio track to determine a preferred transition window
segueing from playing the first track to playing the second track during the preferred transition window.
2. The method of claim 1 , wherein the resampling, comparing and segueing are done automatically without user intervention.
3. The method of claim 1 , wherein comparing the resampled acoustic metadata for the first audio track and the resampled acoustic metadata for the second audio track further comprises identifying and aligning portions of the first track and the second track that sound the most similar.
4. The method of claim 1 , wherein the rhythmic rate of the first track and the rhythmic rate of the second track are the tatum rates for the respective tracks.
5. The method of claim 1 , wherein the acoustic metadata comprises timbral vectors.
6. The method of claim 5 , wherein timbral vectors are calculated by the formula:
V i =W ×( F i −M )
wherein:
V i =timbral vector for the i'th segment of a track
W=a weight matrix calculated from a training data set consisting of a large number of randomly selected tracks
F i =data vector describing the timbral surface of the i'th segment
M=a mean vector calculated from the training data set.
7. The method of claim 6 , wherein the weight matrix is calculated using principle component analysis.
8. The method of claim 1 , wherein performing a sliding-window comparison further comprises
defining a track 2 window including N consecutive samples of the resampled acoustic metadata for the second track
defining a track 1 window including M consecutive samples of the resampled acoustic metadata for the first track
defining a transition portion length of K samples, where K≦M and K≦N
comparing each possible set of K consecutive samples within the track 2 window with each possible set of K consecutive samples within the track 1 window.
9. The method of claim 8 , wherein two sets of K consecutive samples are compared by treating each set of samples as a vector and calculating the Euclidean distance between the two vectors.
10. The method of claim 1 , wherein segueing from playing the first track to playing the second track further comprises:
cross fading the volume of the first track and the second track
time-scaling at least one of the first track and the second track to equalize the tatum rates of the two tracks.
11. The method of claim 1 , wherein segueing from playing the first track to playing the second track further comprises:
smoothly transitioning from the tatum rate of the first track to the tatum rate of the second track.
12. The method of claim 1 , further comprising:
automatically selecting the second track that sounds similar to the first track.
13. The method of claim 12 , wherein automatically selecting the second track further comprises:
comparing acoustic metadata of the first track with acoustic metadata for a plurality of tracks stored in a music library
selecting a second track based on the results of the comparing.
14. The method of claim 13 , wherein comparing acoustic metadata comprises comparing the covariance matrix of the timbral vectors of the first track with the covariance matrices of the timbral vectors for the plurality of tracks.
15. A nontransitory storage medium having instructions stored thereon which, when executed, will cause a computing device to perform actions comprising:
resampling acoustic metadata for a first track at a rhythmic rate of the first track
resampling acoustic metadata for a second track at a rhythmic rate of the second track
performing a sliding window comparison of the resampled acoustic metadata for the first audio track and the resampled acoustic metadata for the second audio track to determine a preferred transition window
segueing from playing the first track to playing the second track during the preferred transition window.
16. The storage medium of claim 15 , wherein the resampling, comparing and segueing are done automatically without user intervention.
17. The storage medium of claim 15 , wherein comparing the resampled acoustic metadata for a first audio track and the resampled acoustic metadata for a second audio track further comprises identifying and aligning portions of the first track and the second track that sound the most similar.
18. The storage medium of claim 15 , wherein the rhythmic rate of the first track and the rhythmic rate of the second track are the tatum rates for the respective tracks.
19. The storage medium of claim 15 , wherein the acoustic metadata comprises timbral vectors.
20. The storage medium of claim 19 , wherein timbral vectors are calculated by the formula:
V i =W ×( F i −M )
wherein:
V i =timbral vector for the i'th segment of a track
W=a weight matrix calculated from a training data set consisting of a large number of randomly selected tracks
F i =data vector describing the timbral surface of the i'th segment
M=a mean vector calculated from the training data set.
21. The storage medium of claim 20 , wherein the weight matrix is calculated using principle component analysis.
22. The storage medium of claim 15 , wherein performing a sliding-window comparison further comprises
defining a track 2 window including N consecutive samples of the the resampled acoustic metadata for the second track
defining a track 1 window including M consecutive samples of the the resampled acoustic metadata for the first track
defining a transition portion length of K samples, where K≦M and K≦N
comparing each possible set of K consecutive samples within the track 2 window with each possible set of K consecutive samples within the track 1 window.
23. The storage medium of claim 22 , wherein two sets of K consecutive samples are compared by treating each set of samples as a vector and calculating the Euclidean distance between the two vectors.
24. The storage medium of claim 15 , wherein segueing from playing the first track to playing the second track further comprises:
cross fading the volume of the first track and the second track
time-scaling at least one of the first track and the second track to equalize the tatum rates of the two tracks.
25. The storage medium of claim 15 , wherein segueing from playing the first track to playing the second track further comprises:
smoothly transitioning from the tatum rate of the first track to the tatum rate of the second track.
26. The storage medium of claim 15 , the actions performed further comprising:
automatically selecting the second track that sounds similar to the first track.
27. The storage medium of claim 26 , wherein automatically selecting the second track further comprises:
comparing acoustic metadata of the first track with acoustic metadata for a plurality of tracks stored in a music library
selecting a second track based on the results of the comparing.
28. The storage medium of claim 27 , wherein comparing acoustic metadata comprises comparing the covariance matrix of the timbral vectors of the first track with the covariance matrices of the timbral vectors for the plurality of tracks.