Manual Reference Pages - GENSGM (1)
gensgm - generate auditory spectrogram
gensgm [ option=value | -option ] [ filename ]
The gensgm module of the AIM software performs a time-domain spectral analysis using a bank of auditory filters, and summarises the information in an auditory spectrogram, that is, a spectrogram with auditory frequency resolution and temporal resolution, rather than the fixed frequency and temporal resolution of traditional speech preprocessors (Patterson and Moore, 1986; Glasberg and Moore, 1990). The spectral analysis converts the input wave into an array of filtered waves, one for each channel of the auditory filterbank. The surface of the array of filtered waves is AIMs representation of basilar membrane motion (BMM) as a function of time (Patterson et al. 1995). The auditory spectrogram is a plot of a sequence of spectral slices extracted from the envelope of the BMM every calculated continuously, by rectifing, compressing, and lowpass filtering the individual BMM waves as they flow from the filterbank.
The frequency resolution of the analysis varies with the center frequency of the channel as in the auditory system, and the distribution of channels across frequency is chosen to match that in the auditory system (Patterson and Moore, 1986). Thus, the auditory spectrogram is a greyscale plot of the activity in each channel (shades of black) as a function of time (the abscissa) and the centre frequency of the auditory filter (the ordinate) in ERBs. The representation is referred to as an auditory spectrogram (SGM) to distinguish it from more traditional spectrograms based on Fourier, LPC or cepstral analysis. In AIM, the suffix sgm is used to distinguish this spectral representation from the other spectral representations provided by the software (asa auditory spectral analysis, cgm cochleogram, and epn excitation pattern).
The spectral analysis performed by gensgm is the same as that performed by genbmm. The primary differences are in the Display defaults and the way the Leaky Integration module is used to produce the spectral slices that form the spectrogram. As a result, this manual entry is restricted to describing the option values that differ from those in genbmm.
I. DISPLAY DEFAULTS
The default values for three of the display options are reset to produce a spectrographic format rather than a landscape. Specifically, display=greyscale, bottom=0 and top=2500. The number of channels is set to 128 for compatibility with the auditory spectrum modules, genasa and genepn. When using AIM as a preprocessor for speech recognition the number of channels would typically be reduced to between 24 and 32. Use option downsample if it is necessary to reduce the output to less than 24 channels across the speech range.
I. RECTIFICATION AND COMPRESSION
The default form of compression is logarithmic; it has the advantage of transforming the exponential envelope of the ringing response of the gammatone filter into a linear decay with time. It also makes the output close to level independant which is useful when using AIM as a preprocessor for speech recognition. There is evidence, however, that auditory compression may be better represented by power compression with an exponent in the range of 0.5. For a discussion of this issue, see docs/aimMeddisHewitt. To accommodate power compression and the assembly of different configurations of AIM, the rectification and compression options are presented separately in the options list before the neural transduction section.
Apply half-wave rectification to filtered waves
Switch. Default value: off.
If rectify is on, the BMM is half-wave rectified. The log compressor also performs half-wave rectification to avoid negative logs. Since the compressor default is log, the rectify default is off.
Note: Full wave rectification is produced if rectify is set to 2. This is useful when calculating envelopes with genasa or gensgm.
Apply compression to filtered waves. The form of the compression can
be either logarithmic (log), or a power function (with a value between
0 and 1).
Switch. Choices log, 0-1, off. Default value: log.
The default compressor is logarithmic, not because it is a particularly good approximation to auditory compression, but rather because it is a good match for the gammatone auditory filter mathematically, and it makes the filterbank level independent. Note that the logarithmic compressor performs half-wave rectification to avoid negative logs.
NOTE: When using the physiological version of AIM with the transmission-line filterbank and the Meddis haircell bank, set compress=off, as compression is an integral part of the feedback loop in the transmission-line filterbank module.
transduction Neural transduction switch (at, meddis, off) Switch. Default: off.
II LEAKY INTEGRATION
stages_idt Number of stages of lowpass filtering Default unit: scalar. Default value: 2 tup_idt The time constant for each filter stage Default unit: ms. Default value: 8 ms.
The Equivalent Rectandular Duration (ERD) of a two stage lowpass filter is about 1.6 times the time constant of each stage, or 12.8 ms in the current case.
downsample The time between successive spectral frames. Default unit: ms. Default value: 10 ms.
Downsample is simply another name for frstep_epn, provided to facilitate a different mode of thinking about time-series data.
frstep_epn The time between successive spectral frames Default unit: ms. Default value: 10 ms.
With a frstep_epn of 10 ms, genasa will produce spectral frames at a rate of 100 per second.
Glasberg, B. R. and B. C. J. Moore (1990). "Derivation of auditory filter shapes from notched-noise data," Hearing Research, 47, 103-138.
Patterson, R.D. and B.C.J. Moore (1986).
"Auditory filters and excitation patterns as representations of
frequency resolution," In: Frequency Selectivity in Hearing. B.C.J.
Moore (Ed.), Academic Press, London. 123-177.
Patterson, R.D., Holdsworth, J. and Allerhand M. (1992a).
"Auditory Models as preprocessors for speech recognition," In: The
Auditory Processing of Speech: From the auditory periphery to words,
M.E.H. Schouten (ed), Mouton de Gruyter, Berlin, 67-83.
Patterson, R.D., Allerhand, M. H. and Holdsworth, J. (1993a).
"Auditory representations of speech sounds," In Visual
representations of speech signals, Eds. Martin Cooke, Steve Beet, and
Malcolm Crawford, John Wiley & Sons, Chichester. 307-314.
Patterson, R.D., Anderson, T., and Allerhand, M. (1994).
"The auditory image model as a preprocessor for spoken language," in
Proc. Third ICSLP, Yokohama, Japan, 1395-1398.
Patterson, R.D., Allerhand, M., and Giguere, C., (1995).
"Time-domain modelling of peripheral auditory processing: A modular
architecture and a software platform," J. Acoust. Soc. Am. 98-3, (in
.gensgmrc The options file for gensgm.
genasa, genbmm, genepn, gencgm
None currently known.
Copyright (c) Applied Psychology Unit, Medical Research Council, 1995
Permission to use, copy, modify, and distribute this software without fee is hereby granted for research purposes, provided that this copyright notice appears in all copies and in all supporting documentation, and that the software is not redistributed for any fee (except for a nominal shipping charge). Anyone wanting to incorporate all or part of this software in a commercial product must obtain a license from the Medical Research Council.
The MRC makes no representations about the suitability of this software for any purpose. It is provided "as is" without express or implied warranty.
THE MRC DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL THE A.P.U. BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
The AIM software was developed for Unix workstations by John Holdsworth and Mike Allerhand of the MRC APU, under the direction of Roy Patterson. The physiological version of AIM was developed by Christian Giguere. The options handler is by Paul Manson. The revised SAI module is by Jay Datta. Michael Akeroyd extended the postscript facilites and developed the xreview routine for auditory image cartoons.
The project was supported by the MRC and grants from the U.K. Defense Research Agency, Farnborough (Research Contract 2239); the EEC Esprit BR Porgramme, Project ACTS (3207); and the U.K. Hearing Research Trust.
|SunOS 5.6||GENSGM (1)||11 May 1995|