TY  - JOUR
T1  - Deep Intelligent System for Human Recognition in Complex Domain
AU - Srivastava, Swati AU - Tripathi, Bipin Kumar 
JO  - Journal of Engineering and Applied Sciences
VL  - 14
IS  - 2
SP  - 373
EP  - 385
PY  - 2019
DA  - 2001/08/19
SN  - 1816-949x
DO  - jeasci.2019.373.385
UR  - https://makhillpublications.co/view-article.php?doi=jeasci.2019.373.385
KW  - Complex neuron structure
KW  -C-TROIKA
KW  -fused fuzzy distribution
KW  -complex neural classifier
KW  -effectiveness
KW  -intelligent system
AB  - This study aims to develop a deep computational model which is a novel aggregation of fuzzy
clustering fused with evolutionary searching and a neural network based on a proposed artificial neuron
structure in complex domain. In our Complex Deep Intelligent System (CDIS), we propose a complex neural
classifier built upon a new complex neuron structure ‘C-TROIKA’. The proposed deep model which is an
amalgamation of Fused Fuzzy Distribution (FFD) and Complex Neural Classifier (CNC) capitulates an efficient
tool for human recognition. The functional aptitudes of conventional neurons have been explored with
complex-valued non-linear aggregation functions. This aggregation has the ability to confine higher-order
correlations among input patterns. The proposed neuron structure based on these aggregation functions
enables the system to provide faster convergence, better learning and recognition accuracy. The effectiveness
and strengths of proposed complex neuron structure ‘C-TROIKA’ based deep intelligent system have been
demonstrated over three benchmark biometric datasets, CASIA iris, Yale face and Indian face to realize the
motivation.
ER  -