Simulation of Noise-Cancelling in the Cockpit of an Aircraft Using Two-Rate Hybrid Neural Network

Abstract

This paper presents the two-rate hybrid neural network (TRHNN) for processing of noisy signal. The received TRHNN consists of "fast" ADAptive LInear NEuron neural network (FADALINENN) and "slow" radial basis neural network (SRBNN). The illustrative design example - noise-cancelling of noisy pilot's voice pattern - was carried out using the TRHNN. The received TRHNN has high speed of signal processing. This example demonstrates that the proposed TRHNN is capable not only to recognize the pilot's voice in the noisy voice pattern, but also to restore the pilot's voice. The simulation results with use the software package Simulink show the computing procedure and applicability of the TRHNNs for fast-acting signal processing and analysis in real-time flight conditions.

Keywords

• multi-rate
• neural networks
• noisy speech
• pattern recognition
• simulation