Browse Prior Art Database

Distributed Coding Pool

IP.com Disclosure Number: IPCOM000018925D
Original Publication Date: 2003-Aug-21
Included in the Prior Art Database: 2003-Aug-21
Document File: 4 page(s) / 18K

Publishing Venue

IBM

Abstract

Disclosed is a system that enables the pooling of manual pattern recognition post processing (e.g. video coding) using a built-in certification mechanism

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 47% of the total text.

Page 1 of 4

Distributed Coding Pool

  Disclosed is a system that enables the pooling of manual pattern recognition post processing (e.g. video coding) using a built-in certification mechanism across the Internet

Traditional pattern recognition in the acoustical / audio and optical/video field
(e.g. speech recognition for audio, OCR or industrial inspection image processing) reaches a certain recognition rate while not exceeding a certain error rate. The rejects have to be coded manually using humans, which are still much better at fuzzy problems like classifying audio or video data than a computer.

    In the case of audio, speech samples along with no or partially recognized ASCII text are being presented to a human operator sitting in front of a PC, who enters or completes the recognition of information, supported by an application, which eases the entry and minimizes the number of key strokes needed.

    In the case of video, images and/or image sequences along with no or partially recognized information are being presented to a human operator sitting in front of a PC, who enters or completes the recognition of information, supported by an application, which eases the entry and minimizes the number of key strokes needed.

    This is being done in each facility/industry having such a need separately. As soon as there are a large number of such requests, there is usually a pool of coders.

Application Examples:

- Video coding of zip codes and streets in postal mail automation environments - Industrial inspection applications, detect visible defects - Surveillance: monitor satellite or traffic scene or building surveillance images - Speech recognition, verify the parts of an audio signal which could not be recognized

This do it yourself approach has the following disadvantages:

- The pool is not fully utilized during non-peak hours - During peak hours, the pool sometimes cannot handle all the requests - Certification of operators is critical and surveillance not possible - Operation and maintenance of such a pool for a singe application is expensive - The pool is usually close to the source of the data having to be coded. This quite often is in a high wage area of the planet.

Solution:

    Faster computing power and higher network bandwidth allow for a distributed coding pool. An audio/video coding service provider could offer coding services to many companies at the same time, fully utilizing his coding pool.. The pool takes the audio/video data of various data sources, dispatches those data to coders somewhere in the world. A coding task switching scheme, an automatic, inherent certification tool, which takes care of training, surveillance, certification, recertification and a fully integrated human resource component (recruiting, training, payment etc.) complete this business method.

Chart giving an overview of the DCP process:

Page 2 of 4

    A customer defines a Service Level Agreement SLA with the DCP service provider, containing the application, expected rec...