Detection and Classification of Noise Using Bark Domain Features
Samrudhi Mohdiwale, Tirath Prasad Sahu, Rahul K. Chaurasia, Naresh Kumar Nagwani, Shrish Verma · 2018
To detect the presence of noise in the speech signal, an amplitude and frequency comparison approach is presented in the paper. This paper also provides a procedure to classify noise into their corresponding type if noise is detected in the signal. A feature called cumulative short time fourier transform has introduced for classification of noise. As the various type of noise can be present in the signal, so a multiclass classification utilized in order to identify the type of noise using support vector machine. The results obtained by proposed method provide better accuracy than classical methods. The paper also contributes to reduce the time complexity of noise classification approach.