F0AC

From STX Wiki
< Programmer Guide‎ | SPU Reference
Revision as of 15:05, 17 May 2010 by 193.171.195.8 (talk) (initial import)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

F0AC - F0 detection using autocorrelation

The autocorrelation function is first applied to all N windowed frames. The maximum number of candidates (MaxNCandidates) are stored for each frame. The result is a MaxNCandidates x N matrix. The cheapest path through this matrix is found using the Viterbi algorithm. The output is a shell table item which contains the f0-values and the center window positions.

The algorithm used is described in detail in: BOERSMA, Paul (1993): Accurate Short-Term Analysis of the Fundamental Frequency and the Harmonics-To-Noise Ratio of a Sampled Sound. Proceedings of the IFA. Vol 17. pp 97-110

Usage:

FOAC WaveItem DeltaT MinimumPitch PeriodsPerWindow MaxNCandidates SilenceThreshold VoicingThreshold OctaveCost OctaveJumpCost VoicedUnvoicedCost MaximumPitch F0Table

Inputs:
WaveItem
The id of the wave item to analyse.
DeltaT
The window shift in seconds (expert parameter!; If 0, then DeltaT is computed automatically by F0AC)
MinimumPitch
The lower pitch boundary in Hz.
PeriodsPerWindow
The number of periods to fit in an analysis window.
MaxNCandidates
The maximum number of autocorrelation candidates for every frame. For every frame a number of f0-candidates (including voiceless candidates) are computed. The atom searches finally one trace through these candidates.
SilenceThreshold
The value for the computation of the strengths of the f0-candidates.
VoicingThreshold
The value for the computation of the strengths of the f0-candidates.
OctaveCost
The value for the computation of the strengths of the f0-candidates.
OctaveJumpCost
The value for the computation of the strengths of the f0-candidates.
VoicedUnvoicedCost
The value for the computation of the strengths of the f0-candidates.
MaximumPitch
The upper pitch boundary in Hz.
F0Table
The shell table item for the result (has to be a 2-column table), column 0 contains f0-values, column 1 the center window positions.
Outputs:
NFrames
The number of frames.
NSampWindow
The number of samples for the analysis window.
DeltaTOut
The DeltaT which was actually used (useful if DeltaT was automatically calculated).
x1
The position of the first frame.