Intelligent audio analysis techniques for identification of music in smart devices. Issue 2 (20th January 2021)

• Record Type: Journal Article
• Authors: Mangla, Pragun; Arora, Shefali; Bhatia, Mohinder Pal Singh
• Is Part Of: Internet technology letters. Volume 5:Issue 2(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-01-20
• Subjects: autoencoders -- CNN -- deep learning -- musical instruments -- RNN -- smart devices -- spoofing; Internet -- Periodicals; 004.67805
• DOI: 10.1002/itl2.268
• Languages: English
• ISSNs: 2476-1508

Abstract

Abstract : Audio classification in smart devices is complex in comparison to image or text classification and recognition as it requires to clean the audio or pre‐process it before extracting its features and thus the accuracy heavily depends upon the pre‐processing of the audio. In this paper, we have used an envelope function having a specific threshold that removes the useless part of the audio which cannot be used to differentiate one audio from another by the smart device. Further, pre‐processed sounds from different musical instruments are subjected to feature extraction and classification using deep convolutional neural networks and Recurrent Neural Networks. It is observed that our proposed approach is suitable for audio classification for smart devices, with results at par with state‐of‐the‐art techniques. Abstract : A deep learning approach to classify sounds from musical instruments. This can be detected by smart devices with the help of Convolutional Neural Networks and Recurrent Neural Networks. Performance us evaluated in terms of metrics like accuracy, precision and recall.