Securing body sensor network (BSN) in an efficient manner is very important for preserving the privacy of medical data. Protecting data confidentiality, integrity and to authenticate the communicating nodes are basic requirements to secure BSN. The existing method to generate entity identifier (EI) from inter-pulse intervals (IPIs) of heartbeats has its advantages in authenticating and identifying nodes, which however was found in this study that such generated EIs are not so resistant to attacks because of potential error patterns. This paper presents an improved scheme of IPI-based EI generation to eliminate the error patterns. The performance of randomness and node identification, i.e. false acceptance rate and false rejection rate, is experimentally evaluated. The results indicate that compared with the existing one, the new scheme is effective to eliminate the error patterns and thus more tolerant to attacks, while there is no compromise on the randomness level and identification performance.