Hello! I am a university undergrad student and I am using Merlin Voice Conversion for my project. I am using the scripts provided in the egs voice conversion folder as my guide. My questions are:
1) I am planning to add Wavelet Transformation to the lf0 features before inputting to the DNN during the training and testing phase. How would I be able to incorporate this idea in the existing codes.
2) Additionally, before the features are inputted to the DNN, I would like to take into account the neighboring frames. I.e. instead of just the features in main frame, I would concatenate features of n frames before and after the main frame into the input vector. I am experimenting if concatenating neighboring frames would improve the performance of the voice conversion.
Hello! I am a university undergrad student and I am using Merlin Voice Conversion for my project. I am using the scripts provided in the egs voice conversion folder as my guide. My questions are:
1) I am planning to add Wavelet Transformation to the lf0 features before inputting to the DNN during the training and testing phase. How would I be able to incorporate this idea in the existing codes.
2) Additionally, before the features are inputted to the DNN, I would like to take into account the neighboring frames. I.e. instead of just the features in main frame, I would concatenate features of n frames before and after the main frame into the input vector. I am experimenting if concatenating neighboring frames would improve the performance of the voice conversion.