The use of electrocardiograms (ECGs) as a modality for biometric recognition has received increasing interest. Whereas ECGs are capable of providing a complete insight into the spatiotemporal nature of the cardiac electrical activity, the large volume of multi-lead recordings makes it challenging to elicit discriminant information therein. Typically, for biometric data to be of use in a recognition task, feature extraction must be performed to remove redundant information and noise from the data and enable the subsequent matching algorithms to work efficiently. In this paper, several feature extraction algorithms for ECG biometric recognition are proposed. Based on the idea of block projection, the proposed algorithms allow the temporal information used by existing single-lead-based techniques to be exploited while taking advantage of the structural information contained in multi-lead ECGs. Besides, these algorithms are applicable to ECGs regardless of their number of leads even to single-lead ones. Like most nonfiducial approaches, they require only one fiducial point (i.e., R peaks) to be determined. Detailed experiments with real data are presented to illustrate the performance of the proposed algorithms.