been considered. The two functions have a support of length
N-1.The features are obtained from the Approximationand Details of the 4
level by using the followingequationsV1=1/d
(Approximation details) (3)Where d = Samples in a frame andV1 = Mean value of approximationV2=1/d
(Approximation or details –V1)) (4)
Where V2=Standard Deviation of approximationV3=maximum (Approximation or details) (5)V4=minimum (Approximation or details) (6)V5=norm (Approximation or Details)
(7)Where V5 = Energy value of frequency3.
.LOCALLY WEIGHTED PROJECTIONREGRESSION (LWPR)LWPR achieves better results in nonlinear functionapproximation in high dimensional spaces. It isinsensitive to redundant data. It uses linear modelslocally [13, 14]. Univariate regressions in selecteddirections are used in the input space. Thenonparametric local learning system learns rapidly. Ituses second order learning methods based onincremental training. Weight adjustments are donebased on local information only. Training LWPR isdone as follows,The 5 features obtained are used as inputs for theLWPR and the target values for training each surfaceroughness type is based on labeling.1.
Input extracted features from wavelet.2.
Initialize LWPR using diagonal distancematrix
, norm, meta rate and initial_
. Manyother variables can be initialized or madeconstants depending upon the requirements.3.
Create random numbers.4.
Choose input and target output of a pattern5.
Find global mean and variance of the patterns.6.
Normalize input and output.7.
Compute the weight.8.
Check if new random field has to be added.9.
Find mean square errors between target andthe estimated values.
Repeat steps 5 to 9 until all the patterns arepresented
Fig.4 Training and testing
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 9, No. 2, February 201143 http://sites.google.com/site/ijcsis/ISSN 1947-5500