An algorithm of fuzzy optimization design based on neural networks is presented as a new method for GPRS congestion control. The combination system adopts a new fuzzy neuron network which can appropriately adjust input and output, and increase robustness, stability and working speed of the network. Besides FNN, wavelet transform also be applied to compression algorithm for a higher and controllable compression rate. As shown by experimental results, the algorithm is capable of significantly preventing the congestion control, which proves that it is a more practical and more effective method than traditional methods for congestion control in wireless networks, and it is especially better at video data transmitting.