Individuals suffering from voice disabilities have limited access to currently available automation technologies that operate through voice commands. To address this issue, an alternative voice recognition approach is essential without directly monitoring the audio signals generated from the vocal cord. In this work, the design of a chemically reactive and conductive sponge is reported to create an underwater vibration sensor with a fast response time and high sensitivity, through orthogonal modulation of conductivity (40–2150 kΩ), water repellence (0°–154°) and mechanical properties (0.32–2.63 MPa). This class of porous sponge sensors enables the identification of subtle water waves generated at the air–water interface and extends its utility to detecting a variety of locomotion (squatting, jumping, walking, etc.), as well as automated voice recognition using a deep learning model without direct contact with the human body. Overall, this underwater vibration sensor provides a novel basis for remote interaction with automated technologies, which finds use in medical diagnostics, human‐machine interfaces, and underwater communication systems.