Heart sound characteristics are linked to blood pressure, and its interpretation is important for detection of cardiovascular disease. In this study, heart sounds' auscultation, acquired from children patients (27 patients, 10.2±3.9 years, 35.7±20.8 kg, 132.3±25.5 cm), were automatically segmented to extract the two main components: the first sound (S1) and the second sound (S2). Following, a set of time, frequency, and wavelet based features, were extracted from the S2, and analyzed in relation to the noninvasive cuff-based measures of blood pressure (mean blood pressure of 78±8.8 mmHg). A multivariate regression analysis was performed for each S2 feature set to determine which features better related to the blood pressure measurements. The best results, in the leave-one-out evaluation, were obtained using the frequency features set, with a MAE of 6.08 mmHg, a MAPE of 7.85%, and a ME of 0.31 mmHg, in the estimation of the mean blood pressure.