FeatureSet type description
% FeatureSet fs
fs
% fs -help
DESCRIPTION
set of features
METHODS
puts
delete delete a feature
name get feature name for a given index
index get feature index for a given name
display displays a feature
SVector insert SVector type object into feature set
FMatrix insert FMatrix type object into feature set
setDesc read a 'Feature Description'
setAccess read a 'File Access Description'
frame return frame of a feature given a featureSet frame index
frameN return featureSet frame number given a list of features
readADC read ADC file
writeADC write ADC file
audioInit init audio device
recordStart start audio recording (see also 'recordGet')
recordGet get new audio data after starting with 'recordStart'
play play audio
adc2pow framebased power
peak framebased peak distance
maxpeak framebased maximum of peak to peak
zero framebased zero crossing rate / sec
silTK T.Kemp's silence feature
spectrum framebased power spectrum
adc2mel 16 framebased melscale coefficients, 8 and 16 kHz only
corr create puls in signals
puls create puls in signals
tone create audio signals
noise create noise signal
xtalk remove crosstalk with an adaptive filter
resample resample audiosignal changing sampling rate
melscale melscale from power spectrum
plp perceptual linear prediction
auditory auditory filterbank
postaud post processing for auditory filterbank
write write feature file
read read feature file
split take coefficients .. of source feature
cut take frames .. of source feature
flip take last frames first
delta symmetrical delta coefficients: x(t+delta) - x(t-delta)
shift shift frames: x(t+delta)
adjacent put adjacent frames together: x(t-delta), x(t+1-delta), ..., x(t+delta)
lin m * source_feature + a
log m * log(source_feature + a)
alog m * log(source_feature + b) with b=max/10^a
exp m * exp(a * source_feature)
pow m * (source_feature ^ a)
maxarg index of maximum value per frame
meanarg mean index per frame
aspike remove spikes from signal
offset remove offset adaptively from signal
filter filter a feature
VTLN Vocal Tract Length Normalization(VTLN)
normframe normalize each frame
mean calculate mean and variance
meansub meansubtraction and variance normalisation
normalize normalize coefficients to range ..
thresh set coefficients to a specified value if they exceed a threshold
beepSeg segment (spectral) feature at beeper positions
silSeg segment (spectral) feature at silence positions
add add two features: a * featureA + b * featureB
mul multiply two features: a * featureA * featureB
matmul multiply matrix A with each frame x of feature: A * x
filterbank multiply band matrix A with each frame x of feature: A * x
concat concat frames (or samples) of features
merge merge coefficients (interleave samples) of features
TCL-DEFINED METHODS
eval run feature description script (featureSetEval)
show show feature set (featshow)
access preprocess feature evaluation parameters (featureSetAccess)
Last modified: Mon Jun 29 16:46:55 EDT 1998