search

tyiannak / pyAudioAnalysis

Python Audio Analysis Library: Feature Extraction, Classification, Segmentation and Applications
Apache License 2.0
5.8k stars 1.19k forks source link

Diarization problems #229

Open giorgionanfa opened 5 years ago

giorgionanfa commented 5 years ago

I observed some errors in the results of the speaker diarization...i don't understand how to solve them. First of all, if i put in input an audio file in which only one person is speaking, i obtain wrong results. Moreover, if i put in input an audio file in which two people are speaking, i obtain the correct result, writing --num 2. At the contrary, if i write --num 0, i obtain that five people are speaking...it is really strange. Can someone help me? I think these bugs should be fixed, also for a simple use of the speaker diarization here. Thanks, Giorgio

Arroosh commented 5 years ago

@ Giorgio I'm also facing these problems. If you know about any other framework having good performance at speaker diarization then please suggest me.

Thanks in advance.

giorgionanfa commented 5 years ago

I tried different frameworks, many of them have some problems...i think that a framework having good performances is the one in the github repository in which i answered to your problem. https://github.com/taylorlu/Speaker-Diarization

Sowhardhhonnappa commented 4 years ago

How did you fix this error for Diarization?

NameError: name 'IOerror' is not defined

Arroosh commented 4 years ago

I think in the pyaudio folder, file name "pyaudio.py" I just edit "IOerror" to "IOError" in it. I'm receiving this error b/c i read wrong file from another directory of pyaudio analysis.

Sowhardhhonnappa commented 4 years ago

Oh ok, let me check it out

Keep getting this, pyAudioAnalysis\audioTrainTest.py", line 520, in load_model_knn except IOerror: NameError: name 'IOerror' is not defined