User-specific learning for improved pedestrian motion modeling in a mobile device

US10323942B2 · kind B2 · utility

Assignee

• QUALCOMM Incorporated · US

Inventors

• Benjamin A. Werner · Sunnyvale, US
• William Morrison · San Francisco, US
• Ning LUO · Cupertino, US
• Ming Sun · Sunnyvale, US
• Joseph Czompo · San Jose, US

Key dates

• Filing date: Sep 23, 2016
• Grant date: Jun 18, 2019
• Expiry date: May 25, 2037

Classification

• Technology area (CPC H): Electricity
• CPC primary: H04W4/38
• WIPO field: Measurement
• WIPO sector: Instruments

Abstract

Techniques provided herein are directed toward enabling on-device learning to create user-specific movement models that can be used for dead reckoning. Because these moving models are user-specific, they can be later used to identify user-specific motions in a manner that provides for a dead reckoning location estimation. In some embodiments, these models can be focused on pedestrian movement, based on the repetitive motion that occurs when a user takes a stride (walking, jogging, running, etc.) or other repetitive motion (swimming, riding a horse, etc.).