DIVIDE

NAME

divide -- divide up binary file using recording prompts file

SYNOPSIS

divide (-l) (-c nchan) (-S spchan) (-L lxchan) (-P promptchan) (-T threshadjust) (-p promptsfile|-n numwords) (-a) -f samprate (-b) (-h headerlen) binfile headfile

DESCRIPTION

divide is a program to break up a binary file into separate SFS files. It is able to cope with multiple channel data and can use prompt file listings from the 'display(?)' program. This program produces a series of clicks which can be recorded on one of the input channels and can be used to synchronise division of the file. Otherwise an energy criterion is used to find the appropriate sections. Energy criteria are only applied to channel 0, on the assumption that this is the speech signal. Waveform outputs are sent to chan0:speech, chan1:lx, chan2:speech, chan3:lx, etc. unless over-rided by the -S and -L switches. The SFS main header information is taken from the supplied file, which is otherwise left unchanged. The token field of each created file is generated from the first word of the lines in the prompts file.

OPTIONS

-I Identify program name and version number.

-l Link output SFS files to binary file, rather than copy.

-c nchan Specify number of channels. Default 1.

-p promptsfile Specify text file containing prompts, one per line. A line ending in '#repeat' is taken to be a corrected version of the previous prompt. The output filename is constructed from the first word on the line, with a numerical extension if the name already exists.

-n numwords Look for this number of items in file. Each is annotated as 'word1', 'word2', etc.

-a Just link and create annotations directly in the SFS file supplied.

-S spchan Specify a channel containing a speech signal [1..N]. The first speech channel specified is also used for energy determination.

-L lxchan Specify the channel containing a laryngographic signal [1..N].

-P promptchan Specify the channel containing the prompting clicks if present [1..N]. This channel is not linked to the final file.

-T threshadjust Allows user to tinker with threshold for word detection. In db, positive value makes less sensitive, negative values more sensitive. Default 0db.

-f samprate Specify the sampling rate of each channel in the binary file. Default 20kHz.

-b Swap bytes in binary file.

-h headerlen Specify length of any header in the binary file to skip in bytes. Default 0.

OUTPUT ITEMS

SP Even channels

LX Odd channels

HISTORY

file input file name

channel input channel number

start start offset (samples)

end end offset (samples)

VERSION/AUTHOR

1.0 Mark Huckvale

SEE ALSO

npoint

BUGS


Fri Jul 09 14:54:42 2004