shuhuagao/vchsm

C++ 11 algorithm implementation for voice conversion using harmonic plus stochastic models

vchsm's Issues

long time in [2]MFCC

First, thanks for sharing this implementation 😄. I utilized the project file in Visual Studio, and I enabled openmp in Language, but when I execute the program, it uses a high level of the CPU, but the time remains slow. I'm making tests on a Intel Core i7 CPU with 8gb RAM. Pardon me if I misunderstood something.

Problems with training datasets

What are the requirements for training data? Must be parallel data?

voice conversion did no effect on ubuntu 12.04

Great thanks for this project, I am doing some study on voice conversion recently.
I complied this project successfully on ubuntu 12.04(64 bit,gcc version 4.9.6), and converted the wav file under test folder. However, even using the model file(Model.dat) you provided, there is no difference between source file and converted file (by my ear). Is there any important thing I missed? Please advise, thanks!

感觉效果不太好。。。

Hi, 作者你好，最近我在调研voice conversion的算法和实现，难得看到一个纯C++的实现，我按照样例测试了下，除了路径配置和Gauss component number之外，没有改其他的参数，demo的效果感觉并不好…我又尝试用voice conversion challenge 2018的公开数据做了一下平行语料的实验，用81段语音测试了下男转女的效果，感觉还是很不好，是我的复现中疏漏了什么参数的配置吗？