Open raniaahmed123 opened 1 year ago
hi bagustris does it mean i should run all python files independently? not just jupyter noteboks? thanks
Yes, you need to generate the feature file (.npy files) by yourself. Hence, it requires the original IEMOCAP dataset.
thank you for your reply :)
NameError: name 'speech' is not defined
when i uncomment this line
NameError: name 'speechs' is not defined
@raniaahmed123
Use this code instead, https://github.com/bagustris/SER_ICSigSys2019/blob/master/code/python_files/save_feature.py
when i uncomment this line #speech = np.delete(speeches, (1061, 1430, 1500, 1552, 1566, 1574, 1575, 1576, 1862, 1863, 1864, 1865, 1868, 1869,
NameError: name 'speechs' is not defined Did you finally get the file using the link given by the author below, I encountered the same problem, using the following code is the same problem, if you solved it, can you reply to me how you solved it, thank you
@noadore, can you explicitly show the error message?
Using the code given in this repository previously, the error makes sense since there is no speech
variable.
The code that I pointed out should work since the speech
variable is defined in line 45.
Afaik, I changed that variable into speeches
when removing silent.
@noadore, can you explicitly show the error message? Using the code given in this repository previously, the error makes sense since there is no
speech
variable. The code that I pointed out should work since thespeech
variable is defined in line 45. Afaik, I changed that variable intospeeches
when removing silent.
I used the above link, reported an error with the above image, and mine has followed the code steps to get data_collected.pickle, thank you for your reply
@noadore,
Check your variable, wheter it is data
or data2
. These code should be run after getting pickle file (data_collected.pickle
).
To obtain the pickle file, you should run mocap_data_collect.py, not the save_feature
above.
So, the flow is: data collection (pickle) --> feature extraction --> classification.
@noadore, Check your variable, wheter it is
data
ordata2
. These code should be run after getting pickle file (data_collected.pickle
).To obtain the pickle file, you should run mocap_data_collect.py, not the
save_feature
above.So, the flow is: data collection (pickle) --> feature extraction --> classification. Thank you very much for your help, I have solved these problems.
I am having a problem in running the save_feature file.
@pranav-iiitdwd The error says cannot find the file. Check your pickle file. Make sure the pickle file is not empty and contains ['signal'] file as the audio waveforms.
My pickle file is approx. 2gb in size, certainly not empty. If possible can you share a snap shot of all the packages and versions you have used, or it would also be helpful if you can host these voiced_feat_files else where. I googled the error, and I think the error is related to pydub package, which is required for audiosegment package.
I give an example in my first comment above (degoo cloud). I write the codes in mid 2019, but I do not remember exactly the version of packages including pydub. Should be the version around that time.
Is it required to add ffmpeg to path, I have added it to path. What else could the error be.
If you used the silence removal method, afaik, there are errors when the threshold is higher than maximum amplitude (all signal are removed). Try to use without silence removal first to see if it works (comment on the second seg
variable in your code above).
Yes, it works without sil removal, the generated file has the exact same size as your file hosted in the first comment. Please give me instructions on how to run, with sil removal. What should be the threshhold, and other things.
Lower the threshold!
threshold = 0.0001
Should works.
I set threshold=0.0001 and duration-=0.1
Still having the same error.
seg = seg.filter_silence(duration_s=duration, threshold_percentage=threshold)
This line gives the error, without this line it works.
Try to lower the duration too, perhaps duration=0.01
.
Also, since there are signals that are lower than thresholds (noisy speech), delete the following instance (or uncomment if you already have it).
## delete noisy speech
speech = np.delete(speeches, (1061, 1430, 1500, 1552, 1566, 1574, 1575, 1576, 1862, 1863, 1864, 1865, 1868, 1869,
1875, 1878, 1880, 1883, 1884, 1886, 1888, 1890, 1892, 1893, 1930, 1931, 1932, 1969,
1970, 1971, 1975, 1976, 1977, 1979, 1980, 1981, 1984, 1985, 1986, 1987, 1988, 1989,
1990, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2002, 2003, 2076, 2106, 2110,
2177, 2178, 2179, 2180, 2206, 2241, 2242, 2243, 2245, 2246, 2253, 2254, 2262, 2263,
2357, 2358, 2359, 2362, 2368, 2373, 2374, 2418, 2523, 2525, 2526, 2534, 2539, 2542,
2549, 2552, 2553, 2554, 2555, 2556, 2561, 2562, 2563, 2564, 2578, 2670, 2671, 2672,
2692, 2694, 2695, 2728, 2733, 2889, 2890, 3034, 3304, 3511, 3524, 3525, 3528, 3655,
3802, 3864, 3930, 4038, 4049, 4051, 4061, 4193, 4241, 4301, 4302, 4307, 4569, 4570), 0)
Reference: https://github.com/bagustris/SER_ICSigSys2019/blob/master/code/python_files/save_feature.py
Tried, still having the same error.
Well, the last way I can suggest is to print the i
index under the for loop. If it fails from the beginning, the version of the package may be the true source of error.
Checked it, it fails from the beginning. Will try checking the package versions. Thanks for the help.
Feel free to post the answer here if you find a solution for the future reader.
I am having a problem in running the save_feature file.
The issue was cleared by installing sox and adding it to path.
Hi @raniaahmed123,
Use this script to generate that file (with adjustment): https://github.com/bagustris/Apsipa2019_SpeechText/blob/master/code/python_files/save_feature.py
I give the example here: https://cloud.degoo.com/share/Ov563dopNnEW14jN_DeBig. But for other files, you need to generate by yourself.