Ballistocardiography (BCG), a measure of body vibrations due to ejection of blood into aorta, has the potential to become a 'zero-effort' cardiovascular health monitoring technology, i.e., a technology that requires little or no engagement on part of the user for its operation. In order for any zero-effort monitoring technology to function without any input from the user, it is important that such a methodology can accurately perform identity recognition and thus continuously provide results and feedback to each user. However, most of the recent work on BCG has focused mainly on the estimation of parameters related to mechanical health and the use of BCG to identify a user has not been explored thoroughly. In this paper, we examine, using discrete cosine transform based features and multi-class linear classifier, the use of BCG heartbeats for identity recognition. We demonstrate from the BCG data of 52 healthy subjects collected using a modified floor tile that an average accuracy of 96.15% can be achieved for correct identification of each subject standing on the tile. Based on these results, we anticipate that such a BCG system, trained for a set of users, can be easily installed at different locations in the house and provide continuous and unobtrusive feedback to users for diagnostic monitoring and quantified-self.