The Main function of wearable device is to accurately detect the object scan by sensor. Strong objective for object recognition SoC is feature key point matching, which is main aim for wearable device application like the mobile application, blind people etc. We address issues including the different performance parameters like the low recognition accuracy, dramatic changes in object view points and time consume by the complex computations to determine the feature points and matching with in-built data library. However, traditional method gives the high computation costs cause long process times. Better performance of SoC ,to achieve high-speed and accurate operation, many methods adopted like scale invariant feature transform (SIFT), gradient matching, large model bases and histograms of receptive field responses and speeded up robust features (SURF) method. This paper give overview of analysis of the improvement in the speed of matching process using HK-Means clustering , self-organizing map and visual vocabulary process (VVP).the performance of given methods and descriptor types in terms of accuracy and speed evaluated on our sign recognition method used for detecting multiple objects in high resolution real time images.