mravanelli / pyspeechrev Goto Github PK

It would be good to keep track of the requirements this project has in a machine readable format such as the common requirement.txt output. By using the requirements specifiers from pip you can then specify versions required for this project, this will make it easier for people to install the dependencies and to use this code.

Hi Mirco,
Thanks for making this tool available. Can you please add some license perhaps MIT license that can allow me to use this repo for some non-academic projects?

Hi, line 42 of the pySpeechRev.py, the normalisation is wrong. You should abs them before u get the max.

From

# Signal normalization
signal_clean=signal_clean/np.abs(np.max(signal_clean))

to

# Signal normalization
signal_clean=signal_clean/np.max(np.abs(signal_clean))

But actually this might not be the best way to normalise a signal. you should probably calulate the power of the signal and scale them according to the power you desire.

Hello Mirco, nice work. I believe it would be better to specify IR's samplerate, and resample it (or the input audio) accordingly before filtering to have meaningful results.
I am assuming you sampled IRs at 16kHz, am I right?

Hi,
The toolkit SincNet is great and in my dataset I got an accuracy of 80% for speaker identification. I read the comments and found that adding reverberation can further improve the accuracy and hence I tried doing that, but when trying to run the script pySpeechRev.py I am getting an error. Can you kindly help me out here.

Hi
what are these IRs under IR_examples? Any document on that?

thanks

Hi
Could you please show me how to create my own IR using ESS technique?
Or could you share the code?

Thank you in advance!

Hello,This code is perfect.I just want to know how do you produce the IR1_16.mat.Is trained from data?Is there any paper I can read?Thank you