Seismic velocity structures have been estimated using ambient noise records observed by distributed acoustic sensing (DAS) technology. Previous studies have obtained S-wave velocity (Vs) structures along submarine cables using surface waves; however P-wave velocity (Vp) structure have seldom been constructed because ambient noise primarily contains surface waves. Here, we show P and Rayleigh wave extractions from ambient seafloor noise observed by DAS, and also estimate the Vp and Vs structures along a submarine cable deployed off Cape Muroto in the Nankai subduction zone, Japan. To extract the P waves, we calculated the cross-correlation functions (CCFs) of the ambient noise and applied a frequency-wavenumber filter to the CCFs to remove the effects of slower surface waves. The results indicate that similar features can be obtained in the Vp and Vs profiles, and that the P wave extractions can be performed on days with poor weather conditions.