Mercurial > hg > emotion-detection-top-level
annotate Code/Descriptors/yin/private/rsmooth.m @ 0:ea0c737c6323
first commit
| author | Dawn Black <dawn.black@eecs.qmul.ac.uk> |
|---|---|
| date | Thu, 26 Jul 2012 14:46:25 +0100 |
| parents | |
| children |
| rev | line source |
|---|---|
| dawn@0 | 1 %y=rsmooth(x,smooth,npasses,trim) - smooth by running convolution |
| dawn@0 | 2 % |
| dawn@0 | 3 % X: input matrix |
| dawn@0 | 4 % SMOOTH: samples - size of square smoothing window |
| dawn@0 | 5 % NPASSES: number of smoothing passes (default=1) |
| dawn@0 | 6 % TRIM: if true, clip Y to same size as X |
| dawn@0 | 7 % |
| dawn@0 | 8 % Y: output matrix |
| dawn@0 | 9 % |
| dawn@0 | 10 % RSMOOTH smooths each column of matrix X by convolution with a square window |
| dawn@0 | 11 % followed by division by the window size. |
| dawn@0 | 12 % Multiple passes allow smoothing with a triangular window (npasses=2), or |
| dawn@0 | 13 % window shapes that approach a gaussian (npasses large). Convolution is |
| dawn@0 | 14 % implemented as a running sum for speed. |
| dawn@0 | 15 % |
| dawn@0 | 16 % Y has NPASSES*(SMOOTH-1) more rows than X unless TRIM is set. |
| dawn@0 | 17 % |
| dawn@0 | 18 % mex function |