The intention of human activity recognition systems is to build a system that can automatically understand a range of predefined activities from the sensor signals. The activity recognition system need to be developed in two fields: feature extraction algorithm from sensor data and classifier. In this study, we present the combination of various time domain features and periodic feature to get optimum feature from the sensor signals. An Artificial Neural Net (ANN) with Scaled Conjugate Gradient (SCG) algorithm has been employed to classify different human activities. For the experiments we constructed a database with five activities such as power walking, jogging, running, standing and walking using a multi-sensor module which is structured as a mobile phone form. Further, two static locations such as hand and pocket were used as a sensor module position. The experimental results showed that the average accuracy of human activity recognition system was about 99.4% from five activities.