nb_params.sr

speech signal process tools

#
# @(#)nb_params.sr	1.2 3/28/90 ESI ATT
# These parameters determine the characteristics of the "narrow-band" 
# spectrogram.  Parameter specifications take the form
#	key_word <value>
# Each line may contain one or several of these pairs.  All specifications
# except the error diffusion filter are optional and may occur in any order.
# All comments and blank lines in this file are optional except the line
# preceeding the spatial filter coefficients which begins with an asterisk.
# Parameters may be changed with any text editor while waves is running so
# that the effects of changes may be readily observed.

# duration of the window applied to the time series (ms).
# The bandwidth of the analysis is inversely proportional to this.
win_size_ms 51.2

# type of time weighting (windowing) to use when computing the DFT's
# 0 ==> rectangular; 1 ==> Hamming; 2 ==> cos**4; 3 ==> Hanning
window_type 1

# time step between adjacent columns in the spectrogram (ms)
# This effects both the time resolution and horizontal magnification.
win_step_ms 2

# coefficient of the first-order preemphasis filter
# A value of -1.0 yields +6dB/oct preemphasis; 0.0 results in no preemphasis.
preemp -.7


# minimum size of the (radix-2) FFT to use in computing spectra (points)
# This is automatically increased as necessary to accommodate the window size.
fft_points	512

# number of pixels displayed vertically for each frequency sample (pixels)
# The vertical size of the display in pixels will thus be
#		(freq_intrp * fft_points)/2.
# This variable (freq_intrp) applies only to spectrograms computed by 
# an external DSP board -- see also min_spec_height in .wave_pro
freq_intrp	2

# factor by which spectral log-magnitude values are multiplied before display
# A value of 1.0 is recommended unless the dynamic range must be limited.
ampl_scale	1

# a constant subtracted from spectral log-magnitude values before display (dB)
# This may be used to adjust for overall signal level, if desired.
ampl_offset_db	50

# fraction of spectral log-magnitude mean subtracted from spectra for display
# Note that this interacts with the ampl_offset_db
agc_ratio	.0

# name of the program to run on the DSP-32 card for spectrogram computation
dsp_prog_name $WAVES_32BIN/dspgm

# fraction of spectral variance used to normalize displayed variance
# This is best left at 0.0 for now...
var_norm	0.0

# dimensions of the error-diffusion spatial filter (pixels)
# These dimensions are only used for monochrome spectrograms.
# For wide-band spectrograms y_dither should be larger than x_dither.
# For narrow-band spectrograms x_dither should be >= y_dither.
# The number of rows and columns should be odd and >= 3.  The coefficients
# may have any value.  If x_dither and y_dither are specified the filter
# coefficients MUST be specified.  The COEFFICIENTS MUST IMMEDIATELY FOLLOW
# THE LINE STARTING WITH THE ASTERISK AND BE THE LAST ITEMS IN THIS FILE.
x_dither	5
y_dither	3

# coefficients of the error-diffusion spatial filter
# The first coefficient in the list is for the lower left corner; the last is
# for the upper right corner.  Note that the last (y_dither/2) coefficients
# are zero and are not used.  The middle entry in the last column is the
# weighting given to the spectral amplitude being quantized.  The other
# coefficients are used to weight the quantization errors of surrounding
# pixels.
* the y_dither by x_dither error-diffusion spatial filter follows
0. 1. 0.
2. 3. 2.
3. 5. 3.
4. 7. 4.
7. 8. 0.