Activity Recognition Based on DKP and OS-KELM Algorithm

This paper presents a systematic design approach for constructing neural classifiers that are capable of classifying human activities using a smartphone. The philosophy of our design approach is to apply first order classifier that separates dynamic activities from static activities and recognizes these two different types of activities separately using second order classifier. We adopt the direct kernel perceptron(DKP)as first classifier for it is a very simple and fast kernel based classifier whose α-coefficients are calculated directly, without any iterative training. And online sequential kernel extreme learning machine (OS-KELM) is chosen as the second classifier for its high efficiency. Experimental results have successfully validated the effectiveness of the proposed recognition scheme.