We introduced a novel non-constrained technique for estimating heart rate variability (HRV) using a ballistocardiogram (BCG). To assess whether the BCG signal can be used to analyse the cardiac autonomic modulation, HRV parameters derived from the BCG signal (ballistocardiographic HRV, B-HRV) were statistically compared with the HRV parameters from the ECG signal during rest and under two different experimental conditions that induce cardiac autonomic rhythm changes: the Valsalva manoeuvre and static exercise. Time domain, frequency domain and nonlinear analyses were individually performed on 15 healthy subjects to assess whether the BCG can be used to analyse the cardiac autonomic modulation under each condition. For all subjects, the proposed method had averages of relative errors of 5.01 ± 4.72, 5.64 ± 4.83 and 5.98 ± 5.80% for resting, Valsalva and post-exercise sessions, respectively, and the correlation coefficients between the reference (ECG) and proposed (BCG) methods are 0.97, 0.98 and 0.98, for resting, Valsalva and post-exercise sessions, respectively. During cardiac autonomic changes, the B-HRV parameters changed in a pattern that is very similar to the variations in the HRV parameters based on Student's t-test results. In addition, some of the B-HRV parameters changed according to cardiac autonomic rhythms controlled by sympathetic and parasympathetic activities during the experiments. These findings indicate that BCG can provide an accurate and reliable means to evaluate autonomic system activation by HRV in its unconstrained way.