Invention Grant
- Patent Title: Dual-precision sensor system using high-precision sensor data to train low-precision sensor data for object localization in a virtual environment
-
Application No.: US16640641Application Date: 2018-08-20
-
Publication No.: US11262837B2Publication Date: 2022-03-01
- Inventor: Matan Protter , Efrat Rotem
- Applicant: Alibaba Technology (Israel) Ltd.
- Applicant Address: IL Herzliya
- Assignee: Alibaba Technology (Israel) Ltd.
- Current Assignee: Alibaba Technology (Israel) Ltd.
- Current Assignee Address: IL Herzliya
- Agency: Pearl Cohen Zedek Latzer Baratz LLP
- International Application: PCT/IL2018/050915 WO 20180820
- International Announcement: WO2019/038757 WO 20190228
- Main IPC: G09G5/00
- IPC: G09G5/00 ; G06F3/01 ; G06N20/00 ; G01P13/00 ; G06N3/08

Abstract:
A system and method is provided for improving sensor precision. In a training phase, a relatively low precision sensor system and a relatively high precision sensor system may record relatively low precision and relatively high precision training motion signals respectively of substantially the same training scene. Machine learning may be performed to generate a transformation mapping information from the relatively low precision training motion signal to the relatively high precision training motion signal. In a run-time phase, a run-time sensor system having a precision significantly less than the precision of the relatively high precision sensor system, may record run-time motion signals of a run-time scene. The run-time motion signal may be transformed using the transformation to generate a transformed run-time motion signal with a precision significantly greater than the precision of the run-time sensor system.
Public/Granted literature
Information query