Diarization of child / adult speech using a Pytorch classifier
Run script is runYunitator.sh
which takes a .wav file as input
and produces in the temp/working folder /vagrant/Yunitemp/
a RTTM
file with matching base name. For example:
vagrant@vagrant-ubuntu-trusty-64:~/Yunitator$ ./runYunitator.sh /vagrant/test2.wav
Extracting features for test2.wav ...
(MSG) [2] in SMILExtract : openSMILE starting!
(MSG) [2] in SMILExtract : config file is: /vagrant/MED_2s_100ms_htk.conf
(MSG) [2] in cComponentManager : successfully registered 95 component types.
(MSG) [2] in cComponentManager : successfully finished createInstances
(19 component instances were finalised, 1 data memories were finalised)
(MSG) [2] in cComponentManager : starting single thread processing loop
(MSG) [2] in cComponentManager : Processing finished! System ran for 1460 ticks.
DONE!
vagrant@vagrant-ubuntu-trusty-64:~/Yunitator$ cat /vagrant/Yunitemp/test2.rttm
SPEAKER test 1 0.0 0.2 <NA> <NA> SIL <NA> <NA>
SPEAKER test 1 1.9 1.6 <NA> <NA> SIL <NA> <NA>
SPEAKER test 1 0.2 1.7 <NA> <NA> CHI <NA> <NA>