Cost efficient active noise cancelling through deep learning classifier and generate alive adder based filter for hearing aids

Abstract

Active noise control (ANC) technology uses sound waves to minimize newlineor eliminate unwanted background noise in a specific region. Numerous newlineresearchers have created various algorithms to improve the quality of speech newlinesignals and reduce noise over the past ten years. The inability to eliminate newlinehigh-frequency noise because of shorter wavelengths, latency problems in newlinereal-time processing, and instability in dynamic situations are some of the newlineproblems that ANC still confronts. In order to tackle these issues, this thesis newlinepresented effective active noise cancellation (ANC) strategies. newlineA Multitude Active Noise cancellation using White Shark Optimized CNN- newlineLSTM Network (MANC Net) has been proposed. Dual tree complex Wavelet newlinetransform is utilized to enhance the quality of audio signal and the signal newlinefeatures are extracted using community detection based Genetic Algorithm. newlineThe interference and desired signals are classified using hybridized CNN- newlineLSTM and the hyper parameters are tuned using White Shark optimization for newlinebetter accuracy. The efficacy of the proposed method is evaluated using newlineaccuracy, specificity, sensitivity, NMSE, STOI and PESQ parameter values in newlinecomparison with other conventional methods. newline