To execute a programme do:
java -cp LIUM_SpkDiarization.jar fr.lium.spkDiarization.programs.<programme name> [options] showname
.
The needed memory size depends of the number of speakers and the duration of the show. For broadcast news, a one hour show with near 30 speakers could be segmented and clustered using less one gigabytes (1g) of memory. By default the size of the JVM memory is very tiny (64MB). To avoid problem please use the JVM option -Xmx<size>
. Where <size>
specifies the maximum size, in bytes, of the memory allocation pool. it could be set in kilobytes, megabytes octet, gigabytes or terabytes by suffixing the value by the letter k,m,g or t (upper case is allowed also).
Help and trace
Help
Except this wiki, the only help is the command line help given by java -cp “LIUM_SpkDiarization.jar” fr.lium.spkDiarization.programs.<programme name> –help
.
for exemple:
java -cp "LIUM_SpkDiarization.jar" fr.lium.spkDiarization.programs.MSegInit --help info[info] ====================================================== info[program] name = MSegInit info[info] ------------------------------------------------------ info[show] [options] show info[ParameterFeature-Input] --fInputMask Features input mask = %s.mfcc info[ParameterFeature-Input] --fInputDesc Features info (type[,s:e:ds:de:dds:dde,dim,c:r:wSize:method]) = sphinx,1:1:0:0:0:0,13,0:0:0:0 info[ParameterFeature-Input] type [sphinx,spro4,gztxt,audio2sphinx] = sphinx (2) info[ParameterFeature-Input] static [0=not present,1=present ,3=to be removed] = 1 info[ParameterFeature-Input] energy [0,1,3] = 1 info[ParameterFeature-Input] delta [0,1,2=computed on the fly,3] = 0 info[ParameterFeature-Input] delta energy [0,1,2=computed on the fly,3] = 0 info[ParameterFeature-Input] delta delta [0,1,2,3] = 0 info[ParameterFeature-Input] delta delta energy [0,1,2,3] = 0 info[ParameterFeature-Input] file dim = 13 info[ParameterFeature-Input] normalization, center [0,1] = 0 info[ParameterFeature-Input] normalization, reduce [0,1] = 0 info[ParameterFeature-Input] normalization, window size = 0 info[ParameterFeature-Input] normalization, method [0 (segment), 1 (cluster), 2 (sliding), 3 (warping)] =0 info[info] ------------------------------------------------------ info[ParameterSegmentationFile-Input] --sInputMask segmentation mask = info[ParameterSegmentationFile-Input ([seg,bck,ctl,saus.seg,seg.xml,media.xml],[ISO-8859-1,UTF8])] --sInputFormat segmentation format = seg,ISO-8859-1 info[ParameterSegmentationFile-Output] --sOutputMask segmentation mask = %s.out.seg info[ParameterSegmentationFile-Output ([seg,bck,ctl,saus.seg,seg.xml,media.xml],[ISO-8859-1,UTF8])] --sOutputFormat segmentation format = seg,ISO-8859-1 info[info] ------------------------------------------------------
Trace
Trace is generated if the option –trace
is given in command line. Most of the trace is prefixed by info[<class name>]
where <class name>
is java class name that generates the trace.
Useful tools
Other set of tools helps to handle the diarization, models or features files. The typical actions applied to diarizations are: filtering, merging, splitting, printing, or extracting.
The tools are stored in package fr.lium.spkDiarization.tools
.