Browse Prior Art Database

Improving Automatic Speech Recognition (ASR) Over Copper Telephone Lines by Exploting Islands of High Quality Large Bandwidth Network Connections

IP.com Disclosure Number: IPCOM000022286D
Original Publication Date: 2004-Mar-04
Included in the Prior Art Database: 2004-Mar-04
Document File: 1 page(s) / 7K

Publishing Venue

IBM

Abstract

Speech recognition degrades over low-quality telephone microphones, and low-bandwidth transmission lines. This inhibits the creation of rich speech automated customer service applications. This invention proposes better exploitation of areas where high quality, high bandwidth transmission is possible. Service organizations can capture customer speech at the locus of their digitial switches, thereby exploiting the higher quality digital telephone microphones and the high bandwidth transmission at these sites. ASR (automatic speech recognition) servers can be placed at these sites, and the speech processing can occur locally. Alternatively, the speech can be digitized locally, and sent as cepstra to the service application server, before speech degradation has occurred.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 61% of the total text.

Page 1 of 1

Improving Automatic Speech Recognition (ASR) Over Copper Telephone Lines by Exploting Islands of High Quality Large Bandwidth Network Connections

Telephone speech recognition is a growing business. Banks, travel agencies etc. provide customer services that are supported by speech recognition instead of (or in addition to) touch-tone automation. One key problem is that speech degrades as it is transmitted over analog lines and this negatively impacts speech recognition accuracy. Also, the standard POTS handsets have lower quality microphones that degrade the signal even before it is transmitted over the telephone lines. Because of these problems, only limited applications that use relatively small vocabularies and limited dialog grammars have flourished. Speech recognition performance is much better, however, if it is transmitted through high quality digital telephone microphones, digital networks or high quality analog lines. Most large companies now have internal digital networks. However, the customer service bureaus that customers call into are connected to these islands of high quality network through limited bandwidth lines that degrade the speech quality. It is necessary to find a solution that allows undegraded speech signals to pass to call centers, despite the prevalence of low bandwidth networks.

The invention suggests installing ASR servers at edges of the islands with good network connections (for example, at the switches that connect the internal dig...