An Improved Multispectral Palmprint Recognition System Using Autoencoder with Regularized Extreme Learning Machine

Algorithm 1

Multispectral palmprint classification based on RELM.

Input: the compressed features of AE for training and testing sets and parameters’ values

Output: the users’ IDs of the testing dataset

Training phase:

(1) Initialization step:

(i) Assign random values for the weights and biases of RELM

(2) Computational step:

(i) Compute the matrix, of the hidden layer using Eq. (11)

(ii) Compute the matrix, of the training set using Eq. (10)

(iii) Compute the output weights, using Eq. (14)

Testing phase:

(1) Computational step:

(i) Compute the matrix, of the hidden layer using Eq. (11)

(ii) Compute the output target value, using Eq. (13)

(2) Classification step:

(i) Classify the testing user’s ID using Eq. (15) depending on whether this ID belongs to the user ID in the training set.