In the ocean, local sound speed variations induce acoustic path changes. Travel-time (TT) variations of acoustic paths are classically used to perform ocean tomography inversion. Initially introduced to cope with multi-arrival interferences and to separate eigenray paths, source-receiver arrays combined with array processing techniques now give access to new observables that could be used for tomography such as direction-of-arrivals (DOAs) and direction of departure (DOD). The cumulative use of TT, DOA, and DOD in the inversion process first requires to study the forward problem which links sound speed variations to these observables measured through array processing from two source-receiver arrays. The so-called sensitivity kernels are established using (1) the first order Born approximation that relates the sound speed variation to the amplitude and phase change of the perturbed received signal and (2) a first order Taylor development which links the received signal perturbations to the relevant observables. In the present work, theoretical TT, DOA, and DOD sensitivity kernels are compared with parabolic equation simulations and tank experiment estimations.