audioaudeering-opensmile

openSMILE: Trying to Extract emotion features from emobase.conf results in error


I was going through the openSMILE book and in section 2.5.6, it mentioned that in order to extract emotion features, one needs to run a command of this sort:

SMILExtract_Release -C config/emobase.conf -I input.wav -O angers.arff -instname ANGER -classes {anger,fear,disgust} -classlabel anger

However, running this command gives an error:

(ERROR) [0] in commandlineParser : doParse: unknown option '-instname' on commandline!

Wanted to know how to fix this. Is -instname a deprecated option? If so, what should it be replaced with?

enter image description here


Solution

  • This is happening because config/emobase.conf doesn't have a definition for instname in the arrfsink component.

    openSMILE allows to define new command line options for the openSMILE binary directly in the configuration file. If you want to define this parameter your config file must have a line like this:

    instanceName=\cm[instname(N){unknown}:instance name]

    You can run opensmile-2.3.0/SMILExtract -h to see which CMD options are available regardless of the configuration file. Other CMD parameters such as -instname should be defined in the configfile. Please check "config\shared\standard_data_output.conf.inc" for an example o how to define this command line option for your configuration file.