People perform several activities during the daily life. It is important to reveal and analyze the daily life characteristic of a person, since it might help to cure several health problems. Especially to overcome obesity, heart attacks etc., people frequently do exercise. However, it is not easy to calculate the consumed energy during these exercises. Extra devices were/are required accomplishing this task. On the other hand, the powerful mobile phones encourage researchers to implement activity recognition task on these smartphones. Thus, activity recognition via mobile phone applications became so popular that several publications are made within the last five years. In this study, we elaborate on the discriminative time-domain features in order to recognize the daily activities with reduced number of features. 70 features, combined from existing studies have been analyzed and 15 of them are selected for the implementation of activity recognition on mobile phone. 6 different classification algorithms and 2 feature selection algorithms have been tested comparatively. The test results show that 8 daily activities including walking, sitting, standing, ascending/descending stairs, jogging, cycling and jumping could be classified with 94% ratio of success rate. Since k-NN is one of the most successful classifier, we have implemented it on our mobile application.