Cross-period quality smoothing in adaptive bitrate algorithm
In some embodiments, a method generates a reward function for an adaptive bitrate function that is based on a bandwidth and profiles in a profile ladder for a current segment of content. The reward function indicates a first profile should be selected for the current segment. The method performs a quality change analysis to select a second profile from the profile. The quality change analysis analyzes a quality change that is based on a previously selected profile from a prior segment and profiles in the profile ladder. The second profile is requested from the profile ladder for the current segment.
1. A method comprising:
calculating, by a computing device, a reward function for an adaptive bitrate function that is based on a bandwidth and profiles in a profile ladder for a current segment of content, wherein the reward function indicates a first profile should be selected for the current segment;
performing, by the computing device, a quality change analysis to select a second profile from the profile ladder, wherein the quality change analysis analyzes a quality change that is based on a previously selected profile from a prior segment and profiles in the profile ladder; and
requesting, by the computing device, the second profile from the profile ladder for the current segment.
2. The method of claim 1 , wherein:
the profile ladder comprises a first profile ladder, and
the prior segment is associated with a second profile ladder.
3. The method of claim 2 , wherein the second profile ladder includes at least one profile with a different bitrate or resolution compared to the bitrate or resolution of profiles in the first profile ladder.
4. The method of claim 2 , wherein the second profile ladder includes a different number of profiles compared to the first profile ladder.
5. The method of claim 2 , wherein:
a third profile in the first profile ladder is associated with the bandwidth, and
the second profile in the second profile ladder is associated with the bandwidth, and
the third profile is associated with a different quality characteristic value than the second profile.
6. The method of claim 5 , wherein the quality characteristic value is a resolution.
7. The method of claim 1 , wherein the reward function comprises a first reward function, and wherein performing the quality analysis comprises:
generating a quality supplement that is based on a quality change from the previously selected profile from the prior segment and profiles of the profile ladder; and
calculating a second reward function that is based on the first reward function and the quality supplement.
8. The method of claim 7 , wherein:
the first reward function assigns first values to profiles in the profile ladder, and
the quality supplement adjusts one or more of the first values of the first reward function to generate second values for the second reward function.
9. The method of claim 8 , wherein the second values are used to select the second profile.
10. The method of claim 9 , wherein:
a highest valued second value is used to select the second profile, and
the second profile is different from a highest valued profile using the first values.
11. The method of claim 7 , wherein:
the quality supplement applies a negative preference to profiles where the quality is reduced from the previously selected profile, and
a quality metric is used to measure a quality reduction between profiles and the previously selected profile.
12. The method of claim 1 , wherein performing the quality analysis comprises:
comparing the first profile to the previously selected profile; and
selecting the second profile based on the comparing.
13. The method of claim 12 , wherein the second profile is selected when a difference in profile levels from the first profile does not meet a threshold.
14. The method of claim 12 , wherein the second profile is selected when a difference in quality characteristic from the first profile does not meet a threshold.
15. The method of claim 12 , wherein:
the second profile is selected when a difference in quality characteristic from the first profile does not meet a first threshold, and
the second profile is selected when a difference in profile levels from the first profile does not meet a second threshold.
16. A non-transitory computer-readable storage medium having stored thereon computer executable instructions, which when executed by a computing device, cause the computing device to be operable for:
calculating a reward function for an adaptive bitrate function that is based on a bandwidth and profiles in a profile ladder for a current segment of content, wherein the reward function indicates a first profile should be selected for the current segment;
performing, a quality change analysis to select a second profile from the profile ladder, wherein the quality change analysis analyzes a quality change that is based on a previously selected profile from a prior segment and profiles in the profile ladder; and
requesting the second profile from the profile ladder for the current segment.
17. A method comprising:
calculating, by a computing device, a first reward function for an adaptive bitrate function that is based on a bandwidth and profiles in a profile ladder for a current segment of content, wherein the first reward function indicates a first profile should be selected for the current segment;
generating, by the computing device, a quality supplement that is based on a quality change from a previously selected profile from a prior segment and profiles of the profile ladder;
generating, by the computing device, a second reward function that is based on the first reward function and the quality supplement, wherein the second reward function indicates a second profile should be selected for the current segment;
comparing, by the computing device, a first span that is associated with a quality characteristic based on the previously selected profile and the second profile;
when the first span does not meet a first threshold, selecting, by the computing device, a third profile that meets the first threshold;
comparing, by the computing device, a second span that is associated with a profile level in the profile ladder based on the previously selected profile and the second profile; and
when the second span does not meet a second threshold, selecting, by the computing device, a fourth profile that meets the first threshold.
18. The method of claim 17 , further comprising:
requesting, by the computing device, one of the second profile, the third profile, and the fourth profile from the profile ladder for the current segment.
19. The method of claim 17 , further comprising:
when the first span meets the first threshold, not selecting the third profile; and
when the second span meets the second threshold, not selecting the fourth profile.
20. The method of claim 19 , wherein:
the second profile is selected when the first span meets the first threshold, and the second span meets the second threshold.