An Adaptive Learning based Generative Adversarial Network for One-To-One Voice Conversion
Sandipan Dhar, Student Member, IEEE, Nanda Dulal Jana, Member, IEEE, and Swagatam Das, Senior Member, IEEE.
Sandipan Dhar, Student Member, IEEE, Nanda Dulal Jana, Member, IEEE, and Swagatam Das, Senior Member, IEEE.
All the results are collected by executing the existing codes of the state-of-the-art models ( CycleGAN-VC2, SP-CycleGAN-VC) and the baseline model (CycleGAN-VC) as well as the proposed ALGAN-VC model on the same machine ( i.e. Dell precision 7820 workstation configured with ubuntu 18.04 64 bit Operating System, Intel Xeon Gold 5215 2.5GHz processor, 96GB RAM, and Nvidia 16GB Quadro RTX5000 graphics ) for the same number of epochs.
Please use headphone to listen to these audio files for a better understanding of the difference in audio quality
Generated Speech Samples for VCC 2016 Speech data set
ALGAN-VC Generated sample
CycleGAN-VC Generated sample
CycleGAN-VC2 Generated sample
SP-CycleGAN-VC Generated sample
Generated Speech Samples for VCC 2018 Speech data set
ALGAN-VC Generated sample
CycleGAN-VC Generated sample
CycleGAN-VC2 Generated sample
SP-CycleGAN-VC Generated sample
Generated Speech Samples for VCC 2020 Speech data set
ALGAN-VC Generated sample
CycleGAN-VC Generated sample
CycleGAN-VC2 Generated sample
SP-CycleGAN-VC Generated sample
Generated Speech Samples for the Self Prepared Indian Regional Language-based Speech data set
ALGAN-VC Generated sample
CycleGAN-VC Generated sample
CycleGAN-VC2 Generated sample
SP-CycleGAN-VC Generated sample