smsAnal

Usage: smsAnal [options] inputSoundFile outputSmsFile

Goto: [ Summary | Description ]

Option Summary

Default values given after each parameter. Click on the boxes after the summaries for more detailed information.

`-`format `1`	analysis formats: 1: harmonic sound 2: inharmonic sound 3: harmonic snd keeping phase info 4: inharmonic snd keeping phase info
`-`stype `0`	input sound types: 0 = multiple notes 1 = single note 2 = single note with no pitch change
`-`direct `0`	analysis direction: 0 = direct (from begining to end) 1 = reverse
`-`detfile name	name of deterministic sound file
`-`resfile name	name of residual sound file
`-`debfile name	name of debug text file
`-`rate `345`	analysis frame rate (frames / second)
`-`swin `3.5`	window size in number of pitch periods
`-`twin `1`	type of window: 0 = Hamming 1 = Blackman-Harris 62dB 2 = Blackman-Harris 70dB 3 = Blackman-Harris 74dB 4 = Blackman-Harris 92dB

`-`lfreq `0`	lowest frequency to be searched
`-`hfreq `12000`	highest frequency to be searched
`-`magt `0.3`	magnitude threshold
`-`dfund `100`	default fundamental
`-`lfund `50`	lowest possible fundamental
`-`hfund `1000`	highest possible fundamental
`-`maxe `20`	maximum fundamental error
`-`freqd `0.45`	maximum frequency deviation
`-`peakc `0.2`	peak contribution to guide
`-`fundc `1`	fundamental contribution to guide
`-`npart `60`	number of partials to be searched for
`-`clean `1`	0 = do not clean partials, 1 = clean partials
`-`minl `0.1`	minimum length of good partials
`-`maxs `0.1`	maximum sleeping time of partials
`-`stoc `1`	0 = no stochastic analysis, 1 = stochastic analysis
`-`ncoeff `16`	number of stochastic coefficients
`-`resflt `500`	cut-off frequency for high-pass filter applied to residual
`-`debug `0`	debug mode: [0 `-` 10]

Description of parameters:

General parameters

-format N (default 1) [1,2,3,4]

format of the analysis representation.

harmonic
inharmonic
harmonic with phase
inharmonic with phase

(formats 3 and 4 cannot be synthesized with smsSynth or smsMk, only with smsSynthDet). Whenever possible use the harmonic representation, it will allow to perform more transformations using smsMk.

-stype N (default 0) [0,1, 2]

type of harmonic sound to be analyzed.

sound phrase
single note
single note with no pitch change. When this is set to 1 the default fundamental (-dfund) is used as the reference fundamental, there is practically no pitch detection. When it is set to 2 there is no pitch detection and the default fundamental is used as fundamental throughout. This is only used with formats 1 and 3, that is, harmonic sounds.
-direct N (default 0) [0,1]
direction of the analysis. 0: direct (from begining to end), 1: reverse. Reverse is useful for percussive sounds or sounds with a noisy attack. When reversed, smsReverse has to be used to get the analysis data in the right direction.
-detfile name (default NIL)
name of deterministic sound file. When given, a sound file with that name will be created with the deterministic sound generated as part of the analysis. This file will only be created when stochastic analysis is performed (-stoc 1). When the analysis has been done with the reverse option (-direct 1) the deterministic file has to be reversed with sndReverse.
-resfile name (default NIL)
name of residual sound file. When given, a sound file with that name will be created with the residual sound generated as part of the analysis. This file will only be created when stochastic analysis is performed (-stoc 1). When the analysis has been done with the reverse option (-direct 1) the residual file has to be reversed with sndReverse.
-debfile name (default NIL)
name of debug txt file. When given, a text file will be created with the original sound, the deterministic synthesis keeping phase and the residual. This file will only be created when stochastic analysis is performed.

STFT parameters

-rate N (default 345) [10 ↔ 1000]

number of frames per second. This will determine the hop size of the analysis window. This value is maintained constant for the whole sound but the window overlapp changes depending on -swin.

-swin R (default 3.5) [3 ↔ 7]

size of analysis window for deterministic analysis expressed as number of pitch periods. The actual window size in seconds will be this value divided by the fundamental frequency found at every given moment. When formats 2 and 4 are used (inharmonic sound) the period is computed from the default fundamental specified (-dfund).

-twin N (default 1) [0,1,2,3,4]

type of analysis window to use in the deterministic analysis:

Hamming
Blackman-Harris 62 dB
Blackman-Harris 70 dB
Blackman-Harris 74 dB
Blackman-Harris 92 dB

Peak detection parameters

-lfreq R (default 0) [0 ↔ 10000]: lowest frequency in Hz of the peaks to be detected. No partials lower than this frequency will be detected. For harmonic sounds this is set to be the same than the lowest fundamental (-lfund).
-hfreq R (default 12000) [10 ↔ 22500]: highest frequency in Hz of the peaks to be detected. No partials higher than this frequency will be detected. It will never be higher than half the sampling-rate.
-magt R (default .3) [0 ↔ ]: minimumspectral magnitude detected, expressed as a power value. No partials softer than this value will be detected. This value will be used as a 0dB.

Pitch detection parameters

-dfund R (default 100) [10 ↔ 5000]: default fundamental frequency in Hz. This is the frequency that is used to set the actual analysis window size when no fundamental has been found by the program. In normal situations it is convenient to give the value of the fundamental frequency of the begining of the sound so that it can start with a good guess. This value has to be between the lowest and the highest fundamental (-lfund and -hfund). In inharmonic sounds this value will be used to set the window size.
-lfund R (default 50) [10 ↔ 5000]: lowest fundamental frequency in Hz to be searched for. No peak is found below this value. Only used in the case of harmonic representation.
-hfund R (default 1000) [10 ↔ 5000]: highest fundamental frequency in Hz to be searched for. Only used in the case of harmonic representation.
-maxe R (default 20) [0 ↔ ]: maximum error to accept a fundamental in a frame. The pitch detection algorithm uses this error as a way to measure the goodness of each fundamental candidate. if it is set very large it will find a pitch at every frame. The higher the value the more tolerant the program is to accept a fundamental. Only used in the case of harmonic representation.

Peak continuation parameters

-freqd R (default .45) [0 ↔ 1]: maximum deviation that is permitted from the "guide frequency" to the continuation peak of the guide, expressed as a percentage. With small values only small changes will be accepted from one frame to the next. There is an automatic correction proportional to frequency so that as the frequency of the guide increases this value decreases.
-peakc R (default .2) [0 ↔ 1]: contribution of the frequency of the previous peak of a given trajectory to the current guide frequency value. If the value is 1, it means that the previous peak will completely define the guide value, the possible current fundamental will not be used to set the guide's frequency. If the value is 0, the previous peak will not be used at all.
-fundc R (default 1) [0 ↔ 1]: contribution of the fundamental frequency of the current frame to the current guide frequency. This is only used in harmonic sounds.
-npart N (default 60) [0 ↔ ]: maximum number of partials to be found in each frame.

Partial cleaning parameters

-clean N (default 1) [0,1]: whether or not to clean the deterministic data after analysis. 0: no cleaning, 1: cleaning. This cleaning process gets rid of short trajectories that may not be part of a stable partial of the sound and also fills gaps according to the values of -minl and -maxs.
-minl R (default .1) [0 ↔ ]: minimum length of the partials in seconds. Partials shorter than this value will be deleted if the -clean flag has been set.
-maxs R (default .1) [0 ↔ ]: maximum sleeping time in seconds for a given partial. Times shorter than this value will be considered gaps in the partial and if the -clean flag has been set, these gaps will be filled by interpolating the boundaries.

Stochastic analysis parameters

-stoc N (default 1) [0,1]: wether the stochastic analysis is done or not. The stochastic analysis is the most time consuming part of the analysis. The first time the analysis is done it is useful to set it to 0 (no analysis), this will let you check if the deterministic analysis was well done, then you can reanalize the sound again incorporating the stochastic analysis.
-ncoeff N (default 16) [2 ↔ 128]: number of coefficients for the stochastic representation. This number corresponds to the number of inflexion points in the spectral envelope.
-resflt R (default 500) [20 ↔ 800]: cut-off frequency of high pass filter applied to residual before stochastic analysis is performed.

Tracing parameters

-debug N (default 0) [0,1,2,3,4,5,6,7,8,9,10]

no debug
debug initialitzation
debug peak detection
debug pitch detection
debug peak continuation
debug cleanning of partials
debug deterministic synthesis
debug stochastic analysis
debug stochastic synthesis
debug top level analysis
debug everything

For an explanation on how all this works, look at the smsAnal.wn document in the docs subdirectory.

craig@ccrma.stanford.edu