Multilayer perceptron (MLP) is a classification of feed forward artificial neural network which utilizes the processing techniques of human reasoning. Multilayer perceptron consists of numerous linear layers of trained network cells called as perceptron along with threshold function. In this paper, we are evaluating particle swarm and multilayer perceptron computational models. The first approach, particle swarm optimization (PSO) algorithm deals with principle components of artificial neural network model to recognize handwritten characters. Handwritten character strokes are collected as feature vectors and based on these vectors, classification of characters takes place. The second approach, bounding box technique is used by MLP (Multi-Layer Perceptron) neural network. The results are compared and their performance and stability in handwritten recognition is evaluated