Browse Prior Art Database

Audio Feedback and Learning for Image Person Recognition Systems

IP.com Disclosure Number: IPCOM000242119D
Publication Date: 2015-Jun-19
Document File: 2 page(s) / 135K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method and system for speech driven feedback and learning for image person recognition is disclosed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

Audio Feedback and Learning for Image Person Recognition Systems

Disclosed is a method and system for speech driven feedback and learning for image person recognition.

For the purposes of this disclosure, a visual person recognition systems is defined as software programs (or suites of software programs) that seek to identify a person based on features extracted from images or video of that person. These systems require user feedback in order to initially train and later improve recognition accuracy. This feedback generally consists of human verification of the recognition system's guesses - usually by providing initial training data (e.g. through some mechanism, informing the computer that "these images are of John Doe; these other images are of Jane Smith; etc..." and asking the system to later identify the person in another image or video). To improve the accuracy of such a system over time, especially in cases where new images are being obtained constantly (such as with overhead Closed-circuit television (CCTV) or security cameras), some sort of continuous feedback can be used, which is then

re-trained into the recognition model to produce a more refined, accurate version. This process may involve human verification of which people are contained in a given image or video. Generally, this type of system requires a user or administrator monitoring the system from a computer or dedicated monitoring device. However, the standard computer based interfaces such as...