In the light of the advantages of non-contact acoustic diagnosis technology, an acoustic image pattern recognizing technique with support vector machine was developed and extended to the field of fault of diagnosis.The noise source recognition and location is carried out by use of the beamforming algorithm for some mechanical fault characteristics. The texture feature and singular value feature were extracted through the acoustic images. Then the training and classification utilizing the support vector machine were processed for the mechanical diagnosis. The good well diagnosis results demonstrate that the technology combining feature extraction with support vector machine can be applied to machinery fault diagnosis and provides a novel attacking approach for the application of acoustic images in fault diagnosis.