Selective Noise Cancellation using Machine Learning

Research on Adaptive Noise Cancellation has concentrated on cancelling all audio signals other than the one the user desires to hear. This paper proposes a method for using machine learning in signal processing to train the model in a manner that can cancel the noise signals. Active noise cancellation has existed for quite some time, but it completely canceled out all the frequencies. But with adaptive noise cancellation, the user should be free to choose the frequencies they want to hear. The goal of this system is to improve and take care of hearing health in noisy cities, such as for example, noise from traffic, engines from factories, reckless driving from drivers, and noise from street vendors, allwhile making sure the important sounds can be still audible by the listener. Previous researches have focused only on training audios based on available datasets, whereas an entire dataset of noise and its features was constructed in order to train the model. Further, a comparison is analysed for the results achieved through the selected algorithms which helps state that KNN provides which a higher accuracy of 86% over Logistic Regression.