In order to effectively improve the accuracy of fiber optic signal recognition, a signal event type discrimination method based on multi classification Gaussian support vector machine (SVM) was adopted. Firstly, the Hanning window convolution method and the principle of 95% energy were used to identify the information of the beginning and end stages of event occurrence. Then, from the perspectives of time domain, frequency domain, and scale domain, the mean and discreteness of various normalized feature parameters were analyzed, and appropriate main feature parameters were selected. Finally, the multi classification Gaussian SVM algorithm was used to classify and recognize three different event types. Theoretical analysis and experimental verification were conducted, and data on different types of fiber optic event signals were obtained. The results showed that pattern recognition of event types in 30 sets of experimental data achieved an accuracy rate of over 96%. This method process meets the high-precision identification requirements of event signals in fiber optic sensing and provides important reference value for the application of fiber optic sensors.