Oates, TimBansal, Naveen2021-01-292021-01-292018-01-0111862http://hdl.handle.net/11603/20715Convolutional Neural Networks (CNNs) have been highly successful in solving computer vision problems like object detection, object recognition, and texture syntheses, but little work has been done on changing speaker identity using CNNs. In this theses, two research areas, namely, neural style transfer and voice conversion, are explored to change speaker identity in audio. Two machine learning models based on convolutional neural networks are proposed and extensive experiments are carried out to evaluate their performance on speech datasets.application:pdfConvolutional Neural NetworkNeural Style TransferVoice ConversionText