ANC system performance analysis using a multiple-LMS-based neural network for high-speed train's noise

Since the eco-era is getting closer, the importance of noise reducing in the passenger cars of high-speed train is very important. The active noise is best choice to reduce low frequency noise because since the passive one too heavy for high speed trains where weight so critical. Also ANC able to reduce the ambient noise when the environmental-factor changes. In this paper, we presents the active noise control (ANC) system based on the least mean square (LMS) algorithm and the neural network algorithm for reducing the interior KTX noise. First, the pure noise of the KTX interior without passengers was measured. Then an LMS framework was constructed as a nominal ANC system, and an artificial single-layered perceptron model was designed as an auxiliary ANC to reduce the real-time residual noise due to the non-stationary and uncertain nature of noise. The parameter vector of the hybrid ANC was determined through online estimation to realize an adaptive ANC configuration by means of the steepest descent algorithm. A simulation experiment was performed to demonstrate the proposed ANC system using the previously mentioned realistic acoustic noise.