Age estimation using neural network

1. Input image convert into gray scale 2. Decompose image using 2-level linear wavelet transformation like haar,symlet and daubechies wavelet transformation. 3. Result of 2 level will be face features like eyes , nose ,chin etc. 2-d is applied using 1-d twice. 4. Distance between face features will be estimated using Euclidean distance and referred as Feature Point Distance(FPD). 5. FPD’s were given input to Neural Network (NN) for classifying age groups. 6. 4 types of FPD: 1. FPD1: distance b/w mid-point of eyes 2. FPD2: distance of mid point of eyes and nose 3. FPD3: eyes and lips 4. FPD4: eyes and chin 7. DE [(i, j), (k, l)]=[ (i-k)^2+ (j-l)^2]^1/2 8. Adaptive Resonance Theory (ART) algorithm is used for classifying age groups. 9. ART is done in 3 steps: 1. Input Unit – F1 layer 2. Cluster Units – F2 layer 3. Reset mechanism to control the similarity of patterns placed on same cluster. 10. 70 face images are given as input for training using ART. 11. Age groups are defined as 1. Child(0 to 12) 2. Adolescence(13 to 18) 3. Adult(19 to 55) 4. Senior adult(56 and above).