A Deep Learning Based System to Predict the Noise (Disturbance) in Audio Files
Kartik P V S M S, Gurusamy Jeyakumar · Advances in parallel computing · 2020
Generally, people prefer their audio to be with very good clarity. They want no disturbances during any interaction and while listening audio files. Automated systems to remove disturbances in an audio file to bring good clarity real time audio communications are in high demand. In this paper a deep learning model to detect the noises in a given audio file is proposed and it working principle is explained. The proposed model was trained, first, to predict the places of noise in the audio file by a well-defined training set which consists of set of audio files with the interval of clear audio and noise. After training, the proposed model predicts the area of disturbance (noise) in any given audio file using the integrated techniques of deep learning and audio processing, and the results are reported. The prediction accuracy of the model was found 90.50 %.