Selecting songs with a desired tempo

1. A method for generating a media playlist of media items matching a target tempo, comprising:

receiving a seed value and a tempo signal from a client device, the tempo signal indicating a target tempo;

calculating a target tempo range associated with the target tempo;

generating a filtered playlist of media items based on the seed value and the target tempo range, wherein the media items in the filtered playlist have a tempo within the target tempo range, wherein generating the filtered playlist comprises:

identifying a media feature tempo for a media item in an initial playlist, wherein the media feature tempo is estimated by a human music analyst,

identifying a machine-learned tempo for the media item in the initial playlist, wherein the machine-learned tempo is estimated by computer analysis,

comparing the media feature tempo and the machine-learned tempo for the media item to determine whether the media feature tempo and the machine-learned tempo match, and

discarding the media item from the initial playlist responsive to the media feature tempo and the machine-learned tempo not matching to generate the filtered playlist; and

providing the media items in the filtered playlist to the client device.

2. The method of claim 1 , wherein generating the filtered playlist comprises:

generating an initial playlist of media items from a library of media items based on the received seed value;

determining tempos of the media items in the initial playlist; and

discarding media items having tempos outside the target tempo range from the initial playlist to generate the filtered playlist.

3. The method of claim 1 , wherein comparing the media feature tempo and the machine-learned tempo comprises:

scaling the machine-learned tempo to produce a scaled machine-learned tempo; and

determining whether the scaled machine-learned tempo matches the media feature tempo.

4. The method of claim 3 , wherein scaling the machine-learned tempo comprises:

scaling the machine-learned tempo by a factor of 2, 3, 4, ½, ⅓, or ¼.

5. The method of claim 1 , wherein calculating the target tempo range comprises:

calculating the target tempo range as a margin below and above the target tempo.

6. The method of claim 1 , wherein the received tempo signal is generated by the client device, and reflects movement of the client device or a user's periodic tapping on the client device.

7. The method of claim 1 , wherein the received tempo signal is generated by identifying characteristic time points in a series of sensor measurements made by the client device, wherein intensity of sensor measurements for the characteristic time points are above a predetermined threshold.

8. The method of claim 7 , wherein the sensor measurements are capacitive sensor measurements.

9. A non-transitory computer-readable storage medium storing executable computer program instructions for generating a media playlist of media items matching a target tempo, the computer program instructions comprising:

receiving a seed value and a tempo signal from a client device, the tempo signal indicating a target tempo;

calculating a target tempo range associated with the target tempo;

generating a filtered playlist of media items based on the seed value and the target tempo range, wherein the media items in the filtered playlist have a tempo within the target tempo range, wherein generating the filtered playlist comprises:

identifying a media feature tempo for a media item in an initial playlist, wherein the media feature tempo is estimated by a human music analyst,

identifying a machine-learned tempo for the media item in the initial playlist, wherein the machine-learned tempo is estimated by computer analysis,

comparing the media feature tempo and the machine-learned tempo for the media item to determine whether the media feature tempo and the machine-learned tempo match, and

discarding the media item from the initial playlist responsive to the media feature tempo and the machine-learned tempo not matching to generate the filtered playlist; and

providing the media items in the filtered playlist to the client device.

10. The computer-readable storage medium of claim 9 , wherein generating the filtered playlist comprises:

generating an initial playlist of media items from a library of media items based on the received seed value;

determining tempos of the media items in the initial playlist; and

discarding media items having tempos outside the target tempo range from the initial playlist to generate the filtered playlist.

11. The computer-readable storage medium of claim 9 , wherein comparing the media feature tempo and the machine-learned tempo comprises:

scaling the machine-learned tempo to produce a scaled machine-learned tempo; and

determining whether the scaled machine-learned tempo matches the media feature tempo.

12. The computer-readable storage medium of claim 11 , wherein scaling the machine-learned tempo comprises:

scaling the machine-learned tempo by a factor of 2, 3, 4, ½, ⅓, or ¼.

13. The computer-readable storage medium of claim 9 , wherein calculating the target tempo range comprises:

calculating the target tempo range as a margin below and above the target tempo.

14. The computer-readable storage medium of claim 9 , wherein the received tempo signal is generated by the client device, and reflects movement of the client device or a user's periodic tapping on the client device.

15. A content server for generating a media playlist of media items matching a target tempo, comprising:

a processor for executing computer program instructions; and

a non-transitory computer-readable storage medium storing computer program instructions executable to perform steps comprising:

receiving a seed value and a tempo signal from a client device, the tempo signal indicating a target tempo;

calculating a target tempo range associated with the target tempo;

generating a filtered playlist of media items based on the seed value and the target tempo range, wherein the media items in the filtered playlist have a tempo within the target tempo range, wherein generating the filtered playlist comprises:

identifying a media feature tempo for a media item in an initial playlist, wherein the media feature tempo is estimated by a human music analyst,

identifying a machine-learned tempo for the media item in the initial playlist, wherein the machine-learned tempo is estimated by computer analysis,

comparing the media feature tempo and the machine-learned tempo for the media item to determine whether the media feature tempo and the machine-learned tempo match, and

discarding the media item from the initial playlist responsive to the media feature tempo and the machine-learned tempo not matching to generate the filtered playlist; and

providing the media items in the filtered playlist to the client device.

16. The content server of claim 15 , wherein generating the filtered playlist comprises:

generating an initial playlist of media items from a library of media items based on the received seed value;

determining tempos of the media items in the initial playlist; and

discarding media items having tempos outside the target tempo range from the initial playlist to generate the filtered playlist.

17. The content server of claim 15 , wherein comparing the media feature tempo and the machine-learned tempo comprises:

scaling the machine-learned tempo to produce a scaled machine-learned tempo; and

determining whether the scaled machine-learned tempo matches the media feature tempo.

18. The content server of claim 17 , wherein scaling the machine-learned tempo comprises:

scaling the machine-learned tempo by a factor of 2, 3, 4, ½, ⅓, or ¼.

19. The content server of claim 15 , wherein calculating the target tempo range comprises:

calculating the target tempo range as a margin below and above the target tempo.

20. The content server of claim 15 , wherein the received tempo signal is generated by the client device, and reflects movement of the client device or a user's periodic tapping on the client device.

21. The method of claim 1 , wherein comparing the media feature tempo and the machine-learned tempo further comprises:

calculating a media feature tempo range as a margin below and above the media feature tempo; and

determining that the media feature tempo for the media item does not match the machine-learned tempo of the media item responsive to determining that the machine-learned tempo or a scaled value of the machine-learned tempo is outside the media feature tempo range.

22. The method of claim 21 , wherein generating the filtered playlist further comprises:

for a second media item in the initial playlist with a second media feature tempo that matches a second machine-learned tempo of the second media item, discarding the second media item from the initial playlist responsive to determining that the second machine-learned tempo or a scaled value of the second machine-learned tempo of the second media item is outside the target tempo range.

23. The computer-readable storage medium of claim 9 , wherein comparing the media feature tempo and the machine-learned tempo further comprises:

calculating a media feature tempo range as a margin below and above the media feature tempo; and

determining that the media feature tempo for the media item does not match the machine-learned tempo of the media item responsive to determining that the machine-learned tempo or a scaled value of the machine-learned tempo is outside the media feature tempo range.

24. The computer-readable storage medium of claim 23 , wherein generating the filtered playlist further comprises:

for a second media item in the initial playlist with a second media feature tempo that matches a second machine-learned tempo of the second media item, discarding the second media item from the initial playlist responsive to determining that the second machine-learned tempo or a scaled value of the second machine-learned tempo of the second media item is outside the target tempo range.

25. The content server of claim 15 , wherein comparing the media feature tempo and the machine-learned tempo further comprises:

calculating a media feature tempo range as a margin below and above the media feature tempo; and

determining that the media feature tempo for the media item does not match the machine-learned tempo of the media item responsive to determining that the machine-learned tempo or a scaled value of the machine-learned tempo is outside the media feature tempo range.

26. The content server of claim 25 , wherein generating the filtered playlist further comprises:

for a second media item in the initial playlist with a second media feature tempo that matches a second machine-learned tempo of the second media item, discarding the second media item from the initial playlist responsive to determining that the second machine-learned tempo or a scaled value of the second machine-learned tempo of the second media item is outside the target tempo range.