Understanding the activity-travel patterns of the traveling public becomes a key to design an efficient transportation infrastructure system. Activity travel information can be obtained from traditional travel surveys, which are very costly and time-consuming. Recently, a large volume of smart phone and cell tower based location data are available from emerging Big Data applications. For example, University of Cambridge developed an Android app, called Device Analyzer, to collect phone use data, including app, phone, wifi usages of the volunteer users. The cell phone location data from such Big Data applications are usually associated with location errors, in a range of 5 meters to 30 meters, or the location data have been randomized or scrambled to protect users’ privacy. A critical data processing challenge is how to use location data with large errors or only partial information to estimate the most likely activity patterns on the transportation network. This paper proposes a new activity pattern identification method that can find the most likely activity types and activity duration at intermediate stops based on the information about the state of the telephony subsystem (i.e. ringing, calling, texting). The proposed algorithm aim to identify home activity duration, major office/work activity duration, rough estimate of departure time and arrival time of trips based on multiday cell phone records from the same user.