In order to effectively extract the fault characteristic signal of bearing and accurate classification, this paper uses the method of introducing the cross validation of genetic algorithm and support vector machine in combination with wavelet packet transformation, to identify the fault bearing issued by the unstable characteristic signal and diagnosis.Firstly, the fault signals of instantaneous changes using wavelet packet transform time-frequency characteristics are extracted. Then, using cross validation of genetic algorithm and support vector machine classifiers are built detection, optimization and fault pattern recognition of parameters. Finally, through the experiment to verify the rationality. The experimental results show that this method has real-time, high accuracy and reliability for the detection and classification of the finite sample fault signal.