Multitude Active Noise cancellation using White Shark Optimized Deep Learning Network

Active noise cancellation (ANC) is an essential feature of audio equipment that reduces unwanted background noise. Unwanted signals in information bearing-signal referred as noise, could degrade the strength of signals in terms of intelligibility and quality. Over the decade, various researchers developed different algorithm to enhance speech signal’s quality and for noise reduction. To address the issue, Multitude Active Noise cancellation using White Shark Optimized CNN-LSTM Network (MANC Net) has been proposed. Initially, Dual tree complex Wavelet transform is utilized to enhance the quality of audio signal with multitude noise and the signal features are extracted using community detection based Genetic Algorithm. Afterwards based on extracted signal, interference and desired signals are classified using hybridized Convolutional neural network - Long short-term memory (CNN-LSTM). Additionally, the hyper parameters of CNN-LSTM are tuned using White Shark optimization for better accuracy. The efficiency of the proposed method is evaluated using accuracy, specificity, sensitivity, NMSE, STOI and PESQ parameter values in comparison with other conventional methods. The higher accuracy rate and low NMSE in classification of audio signals evidenced the efficacy of proposed MANC Net model. The overall accuracy of the proposed is 9.1%, 8.7%, 7.9%, 3.4%, and 1.5% better than FxLMS, deep ANC, CsNNet, MCANC, and GFANC respectively.