signals

speech signal process tools

Signals can be self describing in several dimensions:
	the signal dimentionality
	the form of the stored data
	labels for signal elements
	...

This information might be put in the "description" variable of the header

To handle all data associated with f0 and formant tracking the following
signal types must be supported:
	Periodic
		one-dimensional shorts
			binary
			ascii
		n-dimensional heterogeneous vectors
			ascii
		variable dimension heterogeneous vectors
			ascii
	Aperiodic
		voicing and utterance boundary markers
		hand segmentation data


The raw data files:
	Speech signal (periodic binary shorts)
	F0 signal	(n-dimensional heterogeneous ascii)
		frame number 	(%ld) (for direct readability)
		f0		(%lf)
		energy		(%lf) (in reference window)
		max correlation	(%lf)
		k1		(%lf)
		voicing		(%ld)
	LPC poles	(variable dimensional heterogeneous ascii)
		frame number	(%ld)
		energy		(%lf)	(in preemphasized and windowed data)
		k1		(%lf)	(		"		)
		number of poles	(%ld)	(this many freqs. and bws. follow)
		frequencies	(%lf)
		bandwidths	(%lf)

	Formants	(n-dimensional heterogeneous)
		frame number	(%ld)
		energy from f0	(%lf)	(as in f0 above)
		energy from LPC	(%lf)	(as in poles above)
		max correlation
		k1		(%lf)	(from f0 computation)
		k1		(%lf)	(from pole computation)
		voicing		(%ld)	(from f0 computation)
		formants	(%lf)
		bandwidths	(%lf)