This paper presents a simple method to recognize sign gesturesof American Sign Language using features like number of Peaksand Valleys in an image with its position in an image. Signlanguage is mainly employed by deaf-mutes to communicatewith each other through gestures and visions. We extract theskin part which represents the hand from an image usingL*a*b* Color space. Every hand gesture is cropped from animage such that hand is placed in the center of image for ease offinding features. The system does require hand to be properlyaligned to the camera and does not need any special colormarkers, glove or wearable sensors. The experimental resultsshow that 100% recognition rate for testing and training dataset.