Marine swimmers are able to detect external flow structures and exploit them to reduce locomotive effort. To achieve this advantage, these swimmers use mechanosensory organs to sense their hydrodynamic surroundings. An ability to sense and classify nearby wakes can lead to enhanced control and decision making in the context of bioinspired robotic swimmers and underwater vehicles. In previous work, we used an ideal flow model to demonstrate the viability of detecting and classifying vortex wakes from hydrodynamic sensor measurements. In the present study, we extend our previous wake detection protocol to enable online wake classification from streams of hydrodynamic sensor measurements. The streaming protocol is developed using short-time Fourier analysis and supervised learning algorithms. The online wake detection protocol is found to exhibit a comparable rate of classification accuracy as the original protocol. The online protocol’s performance is also assessed in the face of synthetic measurement noise of various intensities. Noisy measurements degrade the classification accuracy in many cases, but performance can be improved through proper protocol design and tuning.