Recognition of Faces - An Optimized Algorithmic Chain

Face recognition has been one of the popular and important parts in Human Computer Interaction (HCI) systems that find tremendous applications, some of which are very critical like access control, surveillance, etc. There are numerous techniques available to process face images and hence, choosing an optimal algorithmic chain is not a straight forward job. The scenario gets much more interesting while implementing face recognition in applications connected to cloud via Internet of Things (IoT) platform. This paper reviews some of the effective face recognition algorithms and proposes an optimized algorithmic chain offering optimal classification accuracy and lower execution time; thereby making it appropriate for IoT related applications targeting human-centric systems. Also, achieving optimum efficiency by selecting appropriate number of features for a given combination of algorithms and the behaviour of algorithms due to partitioning of the images in case of Local Binary Pattern (LBP) is discussed. Results indicate enhanced classification rates with algorithmic fusion by creating chains or process flow of methods. Accuracy of up to 96% was obtained for one of the chains that were designed. Also, it is evident from the results that this chain outperforms some of the well-known state-of-the-art methods.