The internal noise sources of the bent-axis piston motor are close in distance. For example, the distance between the inlet and outlet of the motor valve plate is 38 mm, and the noise sources have the same frequency and multiple frequency phenomenon. The dense and complex noise sources in the bent-axis motor cause difficulties for the spectrum analysis method to accurately identify the same frequency and multiple frequency signals. The maximum resolution of traditional sound intensity measurement is 50 mm, which cannot meet the requirement of identification accuracy of motor internal noise source. Aiming at the problem that the traditional methods are difficult to identify the motor noise sources accurately, this paper proposed a sound intensity measurement method based on compressed sensing. The compressed sensing theory was applied to the high-precision reconstruction of sound intensity image to obtain the high-resolution sound intensity reconstruction image of the motor. Firstly, the noise radiation simulation of the bent-axis motor was carried out to obtain the characteristics of its external surface sound field. Then, based on the sound intensity image for the motor, a compressed sensing frame applied to the motor sound field was designed to obtain the sound intensity cloud image of the motor with high precision. Finally, the feasibility of the compressed sensing theory to improve the identification accuracy of motor noise source was verified by comparing the traditional acoustic intensity measurement with the compressed sensing acoustic intensity measurement. The results show that the identification scale of motor noise sources are improved from 70 mm to 30 mm by the sound intensity measurement method based on compressed sensing, which improves the accuracy of motor noise sources identification and realizes the high precision location of motor noise sources.