Heart rate variability is an important index to express the tensility and balance of the parasympathetic-sympathetic nerve, and it is valuable to diagnose heart and blood disease in clinical diagnosis. It is approved that HRV signal possess non-linear characteristic, and we can analyse and present HRV signal with non-linear dynamic parameteres, such as fractal dimension. Based on phase-space reconstruction theory, an improved algorithm is introduced to calculate the correlation dimension of HRV signal, and some valuable conclusions are expressed according to the result which is concluded from the correlation dimension of nearly 30 samples.