Analysis of heart rate variability (HRV) with Holter monitoring is often difficult due to excessive artifacts and arrhythmias. While short sudden surges are treated successfully by most methods, slow heart rate (HR) variations, nocturnal trapezoidally-shaped HR increases and special types of arrhythmias which are similar to normal HRV fluctuations may distort further time domain and spectral analysis. This paper examines the advantages and disadvantages of different methods for preprocessing of HR data. We have developed the following approach to the analysis of HRV. (1) A combination method based on the absolute difference between HR values and both the last normal HR value and an updated mean is used for removal of artifacts and arrhythmias. This method can detect both sudden surges in HR values as well as longer periods of noise combined with slow normal variations. An additional stage of wild point removal is then optionally applied. (2) Certain special problems such as large T-waves, bigeminal rhythm, slow HR variations and nocturnal trapezoidally-shaped HR increases are also identified. Although none of the algorithms can be applied successfully to all cases, the final computer analysis for preprocessing described in the present study has proved to be superior to the simplified methods which are usually used and provides more suitable data for HRV analysis.