Development of palm recognition system with machine learning approaches

Palm recognition is one of the vital biometric recognition approaches which is used in personal identification for including a large number of paramount features. Palm image can be used after preprocessing in features extraction process. Features alike palm size, palm skin color has been used as outstanding attributes to recognize the hand palm. In such recognition system, candidates may ask to provide their palm print (whichever right or left palm). Palm prints are usually taken into the system using palm sensor or cameras. In this project, palm images are resourced using a camera after each candidate placed his/her palm inside a pinch mark foe ensuring same orientations and directions for all palm images (candidates). Another attribute that realized in some recognition systems is the utilization of smart sensor to intake the image accurately. Such sensors may have the facility to correct the images (performing a preprocessing) according to its designated configurations. In this project, palm based personal verification system was implemented and based on big palm image dataset that resourced from online image portal (PCOE). However, palm dataset is downloaded from PCOE and used in the further steps of palm recognition system. A right hand palm images are being captured for 112 candidates (males and females). No age rustications were enforced which means all age categories were involved in the study. Hence, palm collection was performed inside PCOE premises as each candidate is asked to place his/her right hand inside a palm place holder. The palm data is further classified using machine learning and deep learning paradigms. Long short term memory neural network has outperformed in prediction the identity base on palm features, LSTM is realized with accuracy of recognition equal to 90%. KEYWORDS:Palm Recognition, Deep Learning, Machine Learning