Hand gesture recognition systems are widely used for Human Computer Interaction (HCI) and sign language recognition. The primary requirement of a hand gesture based application system is to segment the hand/palm part from the other body parts and background in the best possible way. In this paper, we report certain techniques for recognizing isolated English alphabets gestures as well as continuous alphabet gestures. We present an improved segmentation model based on HSV and YCbCr mixed skin-colour space. The classification has been done by a 3 layered Multi-layer Perception Artificial Neural Network (MLP-ANN). Problems such as recognition of similar gestures and movement epenthesis seem to be handled effectively with the proposed techniques.