Contents

This directory contains cepstral files created from the speaker independent portion of the Resource Management (RM1) database. Please check the LDC for details

The material was recorded at 16kHz, with 16-bit resolution, using a Sennheiser HMD-414 headset microphone.

The cepstra files were created with the following arguments, most of them default values from wave2feat:

-nist                  yes
-nchans                1
-whichchan             1
-logspec               no
-feat                  sphinx
-alpha                 0.97
-srate                 16000.0
-frate                 100
-wlen                  0.0256
-nfft                  512
-nfilt                 40
-lowerf                133.33334
-upperf                6855.4976
-ncep                  13
-doublebw              no
-blocksize             200000
-dither                no

The contents of each directory are described individually below.

• feat

The directory has 112 sub-directories, one for each speaker, and containing all of the training and test sets.

• etc

This directory contains the transcriptions, control files, dictionary, phone list, and language model, for both training and test data where appropriate.