Browse Prior Art Database

Echo Suppressor

IP.com Disclosure Number: IPCOM000060875D
Original Publication Date: 1986-May-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 2 page(s) / 35K

Publishing Venue

IBM

Related People

Couturier, C: AUTHOR [+2]

Abstract

In a digital voice network connected to a telephone network, the voice signal is transmitted over a bidirectional 2-wire line to a hybrid transformer which separates incoming from outgoing signals. In conventional echo suppressors, the echo is detected by comparing incoming EL and outgoing ER signal energies within the digital voice processor and opening a switch S whenever ER>EL. In the present application, the process is performed over segments of voice signals 20 ms long. This process presents main drawbacks whenever the local station is talking. A method is proposed here to avoid these drawbacks. The energy ER of the Xout(n) and the energy ER of the Xin(n) signals are computed over each 20 ms block of samples through 10 consecutive values (each 2 ms long, and including 16 samples).

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 94% of the total text.

Page 1 of 2

Echo Suppressor

In a digital voice network connected to a telephone network, the voice signal is transmitted over a bidirectional 2-wire line to a hybrid transformer which separates incoming from outgoing signals. In conventional echo suppressors, the echo is detected by comparing incoming EL and outgoing ER signal energies within the digital voice processor and opening a switch S whenever ER>EL. In the present application, the process is performed over segments of voice signals 20 ms long. This process presents main drawbacks whenever the local station is talking. A method is proposed here to avoid these drawbacks. The energy ER of the Xout(n) and the energy ER of the Xin(n) signals are computed over each 20 ms block of samples through 10 consecutive values (each 2 ms long, and including 16 samples).

(Image Omitted)

The ER information is delayed by a flat delay value FD characterizing the echo path length and then the energies over 20 ms are computed.

(Image Omitted)

Both energy sequences are then integrated into --(k) and --(k) with: --(k) = g . -- (k-1) + (1-g) . ER(k) --(k) = g . -- (k-1) + (1-g) . EL(k) where the constant g<1 governs the integration time. The hybrid gain G is then estimated through G= -- (k)/--(k) Then in operation the current echo suppression decision is taken whenever EL(k) > G . ER(k) The process can be improved by checking that the variations of the ratio EL(k)/ER(k), between current and previous blocks, is more than 20%, i.e., EL(k)/ER(...