An Adaptive Machine Learning Based Speech Enhancement with Background Noise Removal from Multiple Data Corpus

Abstract

In every country, audio noise is a dangerous mechanical toxin that seriously newlinedamages hearing in the workplace. The working people in military, mining, development, newlineprinting, and saw factories tend to lose their hearing performance due to the adverse newlineeffects of noise generated by the machines. They undergo elevated levels of noise, with newlinevarious machinery producing greater levels of noise measured in decibels. The person newlinemay not be able to work under such situations due to serious health issues caused by these newlinesounds. Acoustic disruptions, however, restrict the usage of these applications, deteriorate newlinetheir functionality, or make it harder for the user to appreciate the gadget or understand newlinethe discussion. The communication will not be proper due to the background noise. Few newlinespeech processing algorithm uses single data corpus for processing the speech signal newlineAlgorithms like Least Mean Squares (LMS), Normalized Least Mean Squares newline(NLMS), Filtered x Least Mean Squares (FxLMS) and Filtered x Normalized Least newlineMean Squares (FxNLMS) are frequently being used for noise cancellation. Moreover, newlinethese filters have instability and poor noise reduction; slow convergence also requires a newlinegreater number of filter taps and less performance to identify the unknown system in the newlineActive Noise Canceller (ANC) newline