A method and device for automatically detecting heart valve damage for four heart valves are proposed. The automatic determination method makes use of three or more heart tone microphones to simultaneously record heart tones of a patient's heart, and then separates the heart tones into four heart tone signals of the aortic valve, the pulmonary valve, the tricuspid valve and the mitral valve of the heart based on the timing characteristics and related techniques. Next, these four heart tone signals are digitally processed into sampling signals. Subsequently, the convolution method is used to process the sampling signals for producing system transfer functions. Finally, the system transfer functions and the reference database are compared to verify and determine damage for the four heart valves. The automatic determination method can judge heart valve damage to enhance the quality and convenience of medical treatment.