In this paper, the authors propose a method and system for recognizing everyday human activities by utilizing the acceleration sensing data from the accelerometer instrumented in smart phones. In the method, the human activities are recognized in four steps: data processing, data segmentation, feature extraction, and classification. The proposed method has been deployed on Samsung smart phones and is able to recognize 6 human activities in real-time. the authors rigorously experimented 'on a dataset consisting of 6 everyday activities collected from 17 users using several machine learning algorithms, including support vector machine, Naive Bayesian networks, k-Nearest Neighbors, Decision Tree C4.5, Rule Induction, and Neutral networks. The best accuracies are achieved by Decision Tree C4.5 that demonstrates the human activities can be distinguished with 82 percent precision and 83 percent recall under the leave-one-subject out evaluation protocol. These results have shown the feasibility of smart phone based real-time activity recognition.