PURPOSE: In order to assess whether focal epileptic seizures can be detected and distinguished from exercise we evaluated four different heart rate variability (HRV) methods with short term moving window analysis of 30, 50 or 100 R-R intervals or seconds per analyzed window. METHODS: The four methods consisted of: (1) reciprocal high frequency power based on Fast Fourier Transformation, (2) Cardiac Sympathetic Index (CSI), (3) Modified CSI both based on Lorenz plot, and (4) heart rate differential method. Seventeen patients (12 males, 5 females; age 20-55) had 47 seizures (including three secondary generalized tonic-clonic (sGTC)), which were analyzed during their long term video-EEG monitoring of 1-5 days duration. Positive seizure detection was regarded, when the HRV-value during seizures (1min before to 3min after seizure-onset) exceeded 105% of the highest value during exercise and non-seizure sample-periods of the same patient. RESULTS: Modified CSI100 was the most accurate method: it detected all seizures for 13 of the 17 patients within 6s before till 50s after seizure onset time, even though exercise maximum HR of each patient exceeded that of the seizures. The three sGTC seizures were all detected more than half a minute before the tonic-clonic phase. CONCLUSION: The results indicate a detectable, sudden and inordinate shift toward sympathetic overdrive in the sympathovagal balance of the autonomic nervous system around seizure-onset time, for most patients. The Modified CSI is a promising parameter for a portable ECG-based epilepsy alarm, detecting both focal and sGTC seizures.