GitHub - rosrad/DAE: DAE-models for direverberation in the speech recognition

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 40 Commits
.gitignore		.gitignore
CG_MNIST.m		CG_MNIST.m
README		README
backprop.m		backprop.m
batch_define.m		batch_define.m
calcprop_g.m		calcprop_g.m
conf.base		conf.base
conf.m		conf.m
converter.m		converter.m
dt.conf		dt.conf
filelist.m		filelist.m
generate_task.go		generate_task.go
makebatch.m		makebatch.m
makebatches.m		makebatches.m
minimize.m		minimize.m
mnistdeepauto.m		mnistdeepauto.m
rbm.m		rbm.m
readbatches.m		readbatches.m
task.list		task.list
task.sh		task.sh
task_1.list		task_1.list
translate_mfcc.m		translate_mfcc.m

Repository files navigation

please run as below

converter.m                     #normalize the features
makebatches.m                   #make batches for multi processing
mnistdeepauto.m                 #just like the UBM ,this make a pre tuning (training)
backprop.m                      #fine tuning the DAE model (training)
translate_mfcc.m                #use the DAE model to extract clean sound