In this paper, we recognize Indian Sign Language using gyroscope and accelerometer. The gestures are collected using gyroscope and accelerometer, which is fitted on both the arms of the signer. Gyroscope captures the arm and hand rotation gestures accurately and the accelerometer measures gestures related to vibrations. The obtained gestures are evaluated based on amplitude levels as approved gestures and unapproved gestures.The gesture with high accuracy is extracted from the approved gestures by means of feature extraction technique, where we fix the scale using prior initial values. Min-max scaling method is used in the extraction technique. A particular feature is selected and the selected feature from the dataset is subjected to improved K – means algorithm where clusters are formed. Based on this cluster a classifier is implemented which uses the distance probability technique and thereby the accuracy of the selected feature is found. The algorithm based on inertial sensors produces an accuracy of 96.55% for alphabets and 76.8% for sub words the static gestures are recognized effectively by the hand orientation and improved k-means classifier than the continuous gestures