Learning Pipeline

Step 1 - wav files to htk feature file

General:

HCOPY -C config_filename -S file_with_wav_filenames

note: don't forget to change the output file extension (at the bottom of file_with_wav_filenames)

LPC features:

HCOPY -C lpcconfig.txt -S germanwav.scp
HCOPY -C lpcconfig.txt -S mandarinwav.scp

Step 2 - htk feature file to csv file

htktocsv
example: htktocsvmanyfiles({'features_fbank/mandarin_fbank_files.scp';'features_fbank/german_fbank_files.scp'},'features_fbank_train.csv', 'features_fbank_test.csv')

Step 3 - learn!

open weka explorer
load csv file
discretize last attribute (the class)
classify by picking algorithm and choosing attribute 0 to classify on