No abstract

Adaptive noise cancellation is the process of filtering or estimating a desirable signal from a noise-corrupted observation. This method requires adaptive filters due to the unknown or varying input signal and noise properties. This article provides an overview of the three commonly used adaptive filtering algorithms, namely the Least Mean Square LMS, the Normalized Least Mean Square NLMS and the Recursive Least Square RLS algorithm. These filtering algorithms are investigated to cancel out the noise effect from noise-corrupted data when no reference signal is available. The algorithms are used to elaborate on a specific problem of industrial measurement systems for fault detection, which is corrupted by the thermal noise in sensors. A comparison is shown among these three adaptive algorithms that perform the best desired estimation and noise-canceling effect under the white noise environment. Three comparison criteria are used to evaluate the performance of these algorithms: the error performance, the rate of convergence and the signal-to-noise ratio (SNR). The simulation results demonstrate that the NLMS algorithm shows an excellent noise-cancelling effect in terms of a lower error performance and a higher SNR at the same convergence speed compared to the LMS and RLS algorithm, which improves the fault detection performance of the industrial measurement system.