Title :
An echo canceller based on the multilayer LTBP neural network
Author_Institution :
Beijing Univ. of Posts & Telecommun., China
Abstract :
A novel adaptive echo canceller, based on the multilayer lookup table neural network with error propagation (LTBP) training, is proposed. Using this echo canceller, a perfect echo cancellation and a wider signal dynamic range can be obtained. The LTBP-based echo canceller provides a global minimum and fast convergence, as well as better robustness and fault-tolerance than conventional echo cancellers. The architecture and properties of the LTBP echo canceller are described, and some computer simulations are presented
Keywords :
adaptive filters; backpropagation; echo suppression; neural nets; subscriber loops; table lookup; adaptive echo canceller; architecture; echo canceller; error propagation; multilayer LTBP neural; multilayer lookup table neural network; signal dynamic range; subscriber loops; Adaptive systems; Computer architecture; Convergence; Dynamic range; Echo cancellers; Fault tolerance; Multi-layer neural network; Neural networks; Robustness; Table lookup;
Conference_Titel :
Global Telecommunications Conference, 1992. Conference Record., GLOBECOM '92. Communication for Global Users., IEEE
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-0608-2
DOI :
10.1109/GLOCOM.1992.276602
