This year's PhysioNet challenge consists of locating fetal QRS complexes and estimating QT interval in fetal electrocardiographic (FECG) recordings. The approach presented in this work is based on linear transformations obtained from techniques such as independent (ICA), periodic (πCA) and principal components analysis (PCA). The algorithm consists in a common ECG preprocessing stage, followed by a maternal ECG (MECG) removal, obtaining a preprocessed signal. This preprocessed signal is then rotated by PCA, πCA and ICA, in order to get a more clear representation of the fetal activity. Then QRS complexes are detected in these three signal representations. An statistical model, trained in set A, is applied to the detections obtained from the 12 resulting leads, in order to rank the leads according to their FECG content. With the detections of those leads, an average QRS complex is calculated with the Woody algorithm, and an algorithm for the refinement of the fetal QRS detections (fQRS) is used to end the procedure. An estimation of the SNR gain produced by coherent averaging is used to decide the final validity of the fQRS measurements. For those valid recordings, the QT interval is measured in an average heartbeat, calculated by coherent averaging over all the fQRS. Results in set B were 4714.6 for event 4, and 121.6 for event 5. In conclusion, due to the poor generalization capability of the algorithm, the strategy must be further improved to properly extract FECG.