Cuffless blood pressure (BP) estimation attracts much attention for its convenience and capability of long-term health monitoring. The BP estimation derived from finger photoplethysmography (PPG) and electrocardiographic (ECG) signals and validated by the AAMI protocol is presented in this paper. PPG morphological features, pulse decomposition features, and demographic features were extracted. Combined feature strategy was used for feature compensation and normalization. Interpolation-based oversampling was adopted to improve the balance of training sample distribution. Hierarchical regression is proposed to refine BP estimation by ensemble average and range shrinkage. From the experiments, the mean absolute errors (MAEs) of systolic BP and diastolic BP were 7.55 mmHg and 5.96 mmHg, respectively, with the calibration-free and subject-split criterion. The study first presents the cuffless BP estimation results from PPG and ECG under the AAMI validation protocol.