Invention Grant
- Patent Title: System for determining a predicted buffer condition based on flow metrics and classifier rules generated in response to the creation of training data sets
-
Application No.: US15963354Application Date: 2018-04-26
-
Publication No.: US10778547B2Publication Date: 2020-09-15
- Inventor: Emir Halepovic , Eric Petajan , Vengatanathan Krishnamoorthi , Niklas Carlsson
- Applicant: AT&T Mobility II LLC , AT&T Intellectual Property I, L.P.
- Applicant Address: US GA Atlanta US GA Atlanta
- Assignee: AT&T Intellectual Property I, L.P.,AT&T Mobility II LLC
- Current Assignee: AT&T Intellectual Property I, L.P.,AT&T Mobility II LLC
- Current Assignee Address: US GA Atlanta US GA Atlanta
- Agency: BakerHostetler
- Main IPC: H04L12/26
- IPC: H04L12/26 ; H04L29/08 ; H04L29/06 ; G06N20/00 ; H04L12/24 ; H04L12/851

Abstract:
A buffer classification system calculates first flow metrics for a first flow in response to receiving first packet level data. The system in response to receiving HTTP information, generates buffer emulation data and creates one or more data training sets using the first flow metrics and buffer emulation data, wherein training data used to create the one or more training data sets is further generated by computing an average throughput per second observed over different time windows during a video playback. The system in response to creating the one or more training data sets, generates one or more classifier rules used to determine a buffer condition of a flow. The system in response to receiving second packet level data, calculates second flow metrics for a second flow and predicts a buffer condition for the second flow based on the second flow metrics and the one or more classifier rules.
Public/Granted literature
- US20190334794A1 BUFFER CLASSIFICATION SYSTEM Public/Granted day:2019-10-31
Information query