Design and Implementation of Dual-Microphone Adaptive-Active Noise Cancellation System for De-Noising Speech Signal

摘要

Przedstawiono analizę szumów w otoczeniu przemysłowym i w salach konferencyjnych a następnie przedstawiono metody adaptacyjnych metod redukcji szumów. Szczególną uwagę zwrócono na szum zależny od czasu. Zastosowano metodę podwójnego mikrofonu i wykorzystano sieci neuronowe. Sieć identyfikuje źródło szumu i zakłóceń. Metodę sprawdzono doświadczalnie.%Active Noise Control (ANC) has become an important field of research in recent decades. Noise Control in industrial environments and conference halls as well as in communication systems has been studied under the title adaptive-active noise cancellation-control (AANCC) and the results of these studies have been used in practical applications. Reducing time dependent noise is one of the ways recommended for speech enhancement. Here we have introduced an artificial neural network called ADALINE as a smart dual microphone active noise control system. This artificial neural network identifies sources of noise and interference during its training phase and adjusts accordingly. In this way the system reduces the input signal noise. Tests and implementations presented here are based on speech in Persian language and cumulative white Gaussian noise and the interference is assumed to be of the cosine type.