Initial commit

.gitignore

@@ -0,0 +1,2 @@
+*.pyc
+*.pyo

VAD/AudioInput.conf

@@ -0,0 +1,31 @@
+/*
+This configuration file adds the basic audio input and pre-processing (including different window size ffts)
+*/
+
+
+
+;;;;; this list will be appended to the list in the main config file
+[componentInstances:cComponentManager]
+
+// instance[wave].type=cWaveSource
+instance[wave].type=cPortaudioSource
+
+
+////////////////////////////////////////////////////////////////////////////////////////////////
+// ~~~~~~~~~ Begin of configuration ~~~~~~~~~~~~~~~~~ //////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////
+
+
+//---------------- Audio Recording --------------------------//
+[wave:cPortaudioSource]
+writer.dmLevel=wave
+; audio buffersize must be > then buffersize of framer to avoid 100% cpu hog
+audioBuffersize_sec = 0.1
+buffersize_sec=1.0
+channels=\cm[channels{1}:number of recording channels for live audio]
+sampleRate=\cm[samplerate{16000}: sample-rate for live audio recording]
+listDevices=\cm[listdevices{0}: value 1 = list available portaudio audio devices]
+device=\cm[device{-1}: set portaudio audio device, -1 = default device]
+monoMixdown=1
+
+

VAD/IS13_ComParE_specbase.conf

@@ -0,0 +1,56 @@
+///////////////////////////////////////////////////////////////////////////////////////
+///////// > openSMILE configuration file for ComParE <               //////////////////
+/////////                                                            //////////////////
+/////////  * written 2012 by Florian Eyben *                         //////////////////
+/////////                                                            //////////////////
+///////// (c) 2012 by Florian Eyben                                  //////////////////
+/////////     All rights reserved                                    //////////////////
+///////////////////////////////////////////////////////////////////////////////////////
+
+
+
+;;;;;;; component list ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+[componentInstances:cComponentManager]
+instance[dataMemory].type=cDataMemory
+printLevelStats=0
+
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+[componentInstances:cComponentManager]
+instance[frame25].type=cFramer
+;instance[pe25].type=cVectorPreemphasis
+instance[win25].type=cWindower
+instance[fft25].type=cTransformFFT
+instance[fftmp25].type=cFFTmagphase
+
+[frame25:cFramer]
+reader.dmLevel=wave
+writer.dmLevel=frame25
+frameSize = 0.020
+frameStep = 0.010
+frameCenterSpecial = left
+
+[pe25:cVectorPreemphasis]
+reader.dmLevel=frame25
+writer.dmLevel=frame25pe
+k=0.97
+de=0
+
+[win25:cWindower]
+reader.dmLevel=frame25
+writer.dmLevel=winH25
+winFunc=hamming
+
+[fft25:cTransformFFT]
+reader.dmLevel=winH25
+writer.dmLevel=fftcH25
+
+[fftmp25:cFFTmagphase]
+reader.dmLevel=fftcH25
+writer.dmLevel=fftmagH25
+
+
+
+

VAD/README.txt

@@ -0,0 +1,9 @@
+To run the vad on a wave file:
+
+  sox my.wav -c 1 -s -2 input.wav
+  ./SMILExtract -C Standalone.conf -I input.wav -csv vad.csv
+
+In output_segments/*.wav the voice segments will be stored
+In vad.csv the raw vad activations (-1 to +1) will be dumped. First column is a timestamp in seconds, second column (after the ,) is the vad activation.
+
+

VAD/Standalone.conf

@@ -0,0 +1,49 @@
+
+;;;;; all the components we require are listed here:
+[componentInstances:cComponentManager]
+instance[dataMemory].type=cDataMemory
+ ////// Enable this, to dump the detected turns as wave files (to the same path as SMILExtract.exe is in)
+instance[csvout].type=cCsvSink
+instance[turnOutp].type=cWaveSinkCut
+
+ // set printLevelStats=5 to see the feature names in each level for debugging problems with the feature config
+printLevelStats=3
+nThreads=1
+execDebug=0
+
+
+
+/*
+*************************************************************
+     include configuration files
+*************************************************************
+*/
+
+#### Audio input (look at this file to change the audio sample-rate and sound device, etc.)
+\{WaveFileInput.conf}
+
+####################### VAD configuration (choose one ...) ########################
+
+;; The spectra for the VAD
+\{IS13_ComParE_specbase.conf}
+
+//////// the  LSTM and GM based VAD from the SEMAINE system
+\{lstmVAD.conf}
+
+##################### some config sections in the main file ##########################
+
+;; vad csv output
+[csvout:cCsvSink]
+reader.dmLevel = vad_VAD_voice
+filename = \cm[csv{vad.csv}:vad csv output file]
+printHeader = 0
+timestamp = 1
+number = 0
+delimChar = ,
+
+;;; debug outputs
+[turnOutp:cWaveSinkCut]
+preSil=0.5
+reader.dmLevel = frame25
+fileBase = output_segments/seg
+

VAD/WaveFileInput.conf

@@ -0,0 +1,26 @@
+/*
+This configuration file adds the basic audio input and pre-processing (including different window size ffts)
+*/
+
+
+
+;;;;; this list will be appended to the list in the main config file
+[componentInstances:cComponentManager]
+
+instance[wave].type=cWaveSource
+
+
+////////////////////////////////////////////////////////////////////////////////////////////////
+// ~~~~~~~~~ Begin of configuration ~~~~~~~~~~~~~~~~~ //////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////
+
+
+//---------------- Audio Recording --------------------------//
+[wave:cWaveSource]
+writer.dmLevel=wave
+; audio buffersize must be > then buffersize of framer to avoid 100% cpu hog
+buffersize_sec=1.0
+filename = \cm[inputfile(I){input.wav}:name of wave input file to analyze]
+monoMixdown=1
+
+

VAD/lstmVAD.conf

@@ -0,0 +1,103 @@
+;
+; VOICE ACTIVITY DETECTION CONFIGURATION
+;
+
+[componentInstances:cComponentManager]
+instance[melspec_VAD].type=cMelspec
+instance[plp_VAD].type=cPlp
+instance[delta_VAD].type=cDeltaRegression
+instance[mvn_VAD].type = cVectorMVN
+instance[lstm_vad].type=cRnnProcessor
+instance[dataSelector].type = cDataSelector
+instance[turn].type=cTurnDetector
+
+[melspec_VAD:cMelspec]
+reader.dmLevel=fftmagH25
+writer.dmLevel=melspec_power
+htkcompatible = 0
+nBands = 26
+usePower = 1
+lofreq = 0
+hifreq = 8000
+specScale = mel
+
+[plp_VAD:cPlp]
+reader.dmLevel=melspec_power
+writer.dmLevel=plp_VAD
+buffersize=100
+firstCC = 1
+lpOrder = 18
+cepLifter = 22
+compression = 0.33
+htkcompatible = 0
+newRASTA = 1
+RASTA = 0
+rastaUpperCutoff = 30.0
+rastaLowerCutoff = 6.0
+;rastaUpperCutoff = 29.0
+;rastaLowerCutoff = 0.9
+doIDFT = 1
+doLpToCeps = 1
+doLP = 1
+doInvLog = 1
+doAud = 1
+doLog = 1
+
+[energy_VAD:cEnergy]
+reader.dmLevel=frame25
+writer.dmLevel=energy_VAD
+htkcompatible=1
+rms = 0
+log = 1
+
+[delta_VAD:cDeltaRegression]
+reader.dmLevel=plp_VAD
+writer.dmLevel=plpde_VAD
+deltawin=2
+blocksize=1
+
+[accel_VAD:cDeltaRegression]
+reader.dmLevel=plpde_VAD
+writer.dmLevel=plpdede_VAD
+deltawin=2
+blocksize=1
+
+
+[mvn_VAD:cVectorMVN]
+reader.dmLevel = plp_VAD;plpde_VAD
+writer.dmLevel = plpmvn_VAD
+copyInputName = 1
+processArrayFields = 0
+mode = transform
+initFile = norm.dat
+htkcompatible = 0
+meanEnable = 1
+stdEnable = 1
+normEnable = 0
+
+[lstm_vad:cRnnProcessor]
+reader.dmLevel = plpmvn_VAD
+writer.dmLevel = vad_VAD
+netfile=net.dat
+
+[dataSelector:cDataSelector]
+reader.dmLevel = vad_VAD
+writer.dmLevel = vad_VAD_voice
+nameAppend = vadBin
+copyInputName = 1
+selectedRange = 0
+elementMode = 1
+
+[turn:cTurnDetector]
+reader.dmLevel=vad_VAD_voice
+writer.dmLevel=isTurn
+readVad=1
+threshold = 0.45
+threshold2 = 0.25
+writer.levelconf.noHang=1
+eventRecp = turnOutp
+maxTurnLength=10
+maxTurnLengthGrace=2
+nPre = 11
+nPost = 35
+

VAD/smile.log

@@ -0,0 +1,62 @@
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in SMILExtract : openSMILE starting!
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in SMILExtract : config file is: Standalone.conf
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in cComponentManager : successfully registered 113 component types.
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in configManager : reading config file 'Standalone.conf'
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in configManager : reading config file 'WaveFileInput.conf'
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in configManager : reading config file 'IS13_ComParE_specbase.conf'
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in configManager : reading config file 'lstmVAD.conf'
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in instance 'mvn_VAD' : Loading init file in old MVN binary format
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in smileRnn : Net file format: 2
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in smileRnn : net-task: 1
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in cComponentManager : successfully finished createInstances
+                                 (14 component instances were finalised, 1 data memories were finalised)
+[ 22.06.2014 - 23:13:17 ]
+    (MSG) [2] in cComponentManager : starting single thread processing loop
+[ 22.06.2014 - 23:13:17 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #1
+[ 22.06.2014 - 23:13:17 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #2
+[ 22.06.2014 - 23:13:18 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #4
+[ 22.06.2014 - 23:13:18 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #5
+[ 22.06.2014 - 23:13:18 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #9
+[ 22.06.2014 - 23:13:19 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #23
+[ 22.06.2014 - 23:13:20 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #25
+[ 22.06.2014 - 23:13:20 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #27
+[ 22.06.2014 - 23:13:20 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #29
+[ 22.06.2014 - 23:13:20 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #30
+[ 22.06.2014 - 23:13:20 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #33
+[ 22.06.2014 - 23:13:20 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #35
+[ 22.06.2014 - 23:13:21 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #36
+[ 22.06.2014 - 23:13:21 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #43
+[ 22.06.2014 - 23:13:22 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #45
+[ 22.06.2014 - 23:13:22 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #49
+[ 22.06.2014 - 23:13:22 ]
+    (ERROR) [1] in instance 'turnOutp' : no frames were written for turn #52
+[ 22.06.2014 - 23:13:22 ]
+    (MSG) [2] in cComponentManager : Processing finished! System ran for 25572 ticks.
+here in destr.